best-practices/feature-engineering/feature_engineering.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Best Practices - Feature Engineering\n",
    "\n",
    "This notebook walks through recommended feature engineering techniques that can be used to improve model performance.\n",
    "\n",
    "**Steps**\n",
    "\n",
    "1. Import and clean data\n",
    "2. Train baseline model\n",
    "3. Convert text features to word-embeddings\n",
    "4. Perform target encoding on categorical features\n",
    "5. Add anomaly score\n",
    "6. Evaluate improvement to model\n",
    "7. Analyze final model\n",
    "\n",
    "## Step 1. Import and Clean Data\n",
    "\n",
    "We will start by importing the raw lending club data into our H2O-3 cluster."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Checking whether there is an H2O instance running at http://localhost:54321..... not found.\n",
      "Attempting to start a local H2O server...\n",
      "  Java Version: java version \"1.8.0_91\"; Java(TM) SE Runtime Environment (build 1.8.0_91-b14); Java HotSpot(TM) 64-Bit Server VM (build 25.91-b14, mixed mode)\n",
      "  Starting server from /Users/megankurka/anaconda3/lib/python3.6/site-packages/h2o/backend/bin/h2o.jar\n",
      "  Ice root: /var/folders/gv/w2f3zs_d33l3dt5j67k9nhhr0000gn/T/tmpeikasb1m\n",
      "  JVM stdout: /var/folders/gv/w2f3zs_d33l3dt5j67k9nhhr0000gn/T/tmpeikasb1m/h2o_megankurka_started_from_python.out\n",
      "  JVM stderr: /var/folders/gv/w2f3zs_d33l3dt5j67k9nhhr0000gn/T/tmpeikasb1m/h2o_megankurka_started_from_python.err\n",
      "  Server is running at http://127.0.0.1:54321\n",
      "Connecting to H2O server at http://127.0.0.1:54321... successful.\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<div style=\"overflow:auto\"><table style=\"width:50%\"><tr><td>H2O cluster uptime:</td>\n",
       "<td>01 secs</td></tr>\n",
       "<tr><td>H2O cluster timezone:</td>\n",
       "<td>America/Los_Angeles</td></tr>\n",
       "<tr><td>H2O data parsing timezone:</td>\n",
       "<td>UTC</td></tr>\n",
       "<tr><td>H2O cluster version:</td>\n",
       "<td>3.22.0.1</td></tr>\n",
       "<tr><td>H2O cluster version age:</td>\n",
       "<td>12 days </td></tr>\n",
       "<tr><td>H2O cluster name:</td>\n",
       "<td>H2O_from_python_megankurka_eufsnh</td></tr>\n",
       "<tr><td>H2O cluster total nodes:</td>\n",
       "<td>1</td></tr>\n",
       "<tr><td>H2O cluster free memory:</td>\n",
       "<td>3.556 Gb</td></tr>\n",
       "<tr><td>H2O cluster total cores:</td>\n",
       "<td>8</td></tr>\n",
       "<tr><td>H2O cluster allowed cores:</td>\n",
       "<td>8</td></tr>\n",
       "<tr><td>H2O cluster status:</td>\n",
       "<td>accepting new members, healthy</td></tr>\n",
       "<tr><td>H2O connection url:</td>\n",
       "<td>http://127.0.0.1:54321</td></tr>\n",
       "<tr><td>H2O connection proxy:</td>\n",
       "<td>None</td></tr>\n",
       "<tr><td>H2O internal security:</td>\n",
       "<td>False</td></tr>\n",
       "<tr><td>H2O API Extensions:</td>\n",
       "<td>XGBoost, Algos, AutoML, Core V3, Core V4</td></tr>\n",
       "<tr><td>Python version:</td>\n",
       "<td>3.6.0 final</td></tr></table></div>"
      ],
      "text/plain": [
       "--------------------------  ----------------------------------------\n",
       "H2O cluster uptime:         01 secs\n",
       "H2O cluster timezone:       America/Los_Angeles\n",
       "H2O data parsing timezone:  UTC\n",
       "H2O cluster version:        3.22.0.1\n",
       "H2O cluster version age:    12 days\n",
       "H2O cluster name:           H2O_from_python_megankurka_eufsnh\n",
       "H2O cluster total nodes:    1\n",
       "H2O cluster free memory:    3.556 Gb\n",
       "H2O cluster total cores:    8\n",
       "H2O cluster allowed cores:  8\n",
       "H2O cluster status:         accepting new members, healthy\n",
       "H2O connection url:         http://127.0.0.1:54321\n",
       "H2O connection proxy:\n",
       "H2O internal security:      False\n",
       "H2O API Extensions:         XGBoost, Algos, AutoML, Core V3, Core V4\n",
       "Python version:             3.6.0 final\n",
       "--------------------------  ----------------------------------------"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "import h2o\n",
    "h2o.init()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Parse progress: |█████████████████████████████████████████████████████████| 100%\n"
     ]
    }
   ],
   "source": [
    "data_path = \"https://s3.amazonaws.com/h2o-public-test-data/bigdata/laptop/lending-club/LoanStats3a.csv\"\n",
    "data = h2o.import_file(data_path, destination_frame = \"loan_raw\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<table>\n",
       "<thead>\n",
       "<tr><th style=\"text-align: right;\">         id</th><th style=\"text-align: right;\">  member_id</th><th style=\"text-align: right;\">  loan_amnt</th><th style=\"text-align: right;\">  funded_amnt</th><th style=\"text-align: right;\">  funded_amnt_inv</th><th>term     </th><th>int_rate  </th><th style=\"text-align: right;\">  installment</th><th>grade  </th><th>sub_grade  </th><th>emp_title                </th><th>emp_length  </th><th>home_ownership  </th><th style=\"text-align: right;\">  annual_inc</th><th>verification_status     </th><th>issue_d            </th><th>loan_status  </th><th>pymnt_plan  </th><th>url                                                                 </th><th>desc                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                        </th><th>purpose           </th><th>title                                </th><th>zip_code  </th><th>addr_state  </th><th style=\"text-align: right;\">  dti</th><th style=\"text-align: right;\">  delinq_2yrs</th><th>earliest_cr_line   </th><th style=\"text-align: right;\">  inq_last_6mths</th><th>mths_since_last_delinq  </th><th>mths_since_last_record  </th><th style=\"text-align: right;\">  open_acc</th><th style=\"text-align: right;\">  pub_rec</th><th style=\"text-align: right;\">  revol_bal</th><th>revol_util  </th><th style=\"text-align: right;\">  total_acc</th><th>initial_list_status  </th><th style=\"text-align: right;\">  out_prncp</th><th style=\"text-align: right;\">  out_prncp_inv</th><th style=\"text-align: right;\">  total_pymnt</th><th style=\"text-align: right;\">  total_pymnt_inv</th><th style=\"text-align: right;\">  total_rec_prncp</th><th style=\"text-align: right;\">  total_rec_int</th><th style=\"text-align: right;\">  total_rec_late_fee</th><th style=\"text-align: right;\">  recoveries</th><th style=\"text-align: right;\">  collection_recovery_fee</th><th>last_pymnt_d       </th><th style=\"text-align: right;\">  last_pymnt_amnt</th><th>next_pymnt_d       </th><th>last_credit_pull_d  </th><th style=\"text-align: right;\">  collections_12_mths_ex_med</th><th>mths_since_last_major_derog  </th><th style=\"text-align: right;\">  policy_code</th></tr>\n",
       "</thead>\n",
       "<tbody>\n",
       "<tr><td style=\"text-align: right;\">1.0775e+06 </td><td style=\"text-align: right;\">1.2966e+06 </td><td style=\"text-align: right;\">       5000</td><td style=\"text-align: right;\">         5000</td><td style=\"text-align: right;\">             4975</td><td>36 months</td><td>10.65%    </td><td style=\"text-align: right;\">       162.87</td><td>B      </td><td>B2         </td><td>                         </td><td>10+ years   </td><td>RENT            </td><td style=\"text-align: right;\">       24000</td><td>VERIFIED - income       </td><td>2011-12-01 00:00:00</td><td>Fully Paid   </td><td>n           </td><td>https://www.lendingclub.com/browse/loanDetail.action?loan_id=1077501</td><td>Borrower added on 12/22/11 > I need to upgrade my business technologies.<br>                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                </td><td>credit_card       </td><td>Computer                             </td><td>860xx     </td><td>AZ          </td><td style=\"text-align: right;\">27.65</td><td style=\"text-align: right;\">            0</td><td>1985-01-01 00:00:00</td><td style=\"text-align: right;\">               1</td><td>                        </td><td>                        </td><td style=\"text-align: right;\">         3</td><td style=\"text-align: right;\">        0</td><td style=\"text-align: right;\">      13648</td><td>83.7%       </td><td style=\"text-align: right;\">          9</td><td>f                    </td><td style=\"text-align: right;\">       0   </td><td style=\"text-align: right;\">           0   </td><td style=\"text-align: right;\">      5861.07</td><td style=\"text-align: right;\">          5831.78</td><td style=\"text-align: right;\">          5000   </td><td style=\"text-align: right;\">         861.07</td><td style=\"text-align: right;\">                0   </td><td style=\"text-align: right;\">        0   </td><td style=\"text-align: right;\">                     0   </td><td>2015-01-01 00:00:00</td><td style=\"text-align: right;\">           171.62</td><td>                   </td><td>2015-01-01 00:00:00 </td><td style=\"text-align: right;\">                           0</td><td>                             </td><td style=\"text-align: right;\">            1</td></tr>\n",
       "<tr><td style=\"text-align: right;\">1.07743e+06</td><td style=\"text-align: right;\">1.31417e+06</td><td style=\"text-align: right;\">       2500</td><td style=\"text-align: right;\">         2500</td><td style=\"text-align: right;\">             2500</td><td>60 months</td><td>15.27%    </td><td style=\"text-align: right;\">        59.83</td><td>C      </td><td>C4         </td><td>Ryder                    </td><td>< 1 year    </td><td>RENT            </td><td style=\"text-align: right;\">       30000</td><td>VERIFIED - income source</td><td>2011-12-01 00:00:00</td><td>Charged Off  </td><td>n           </td><td>https://www.lendingclub.com/browse/loanDetail.action?loan_id=1077430</td><td>Borrower added on 12/22/11 > I plan to use this money to finance the motorcycle i am looking at. I plan to have it paid off as soon as possible/when i sell my old bike. I only need this money because the deal im looking at is to good to pass up.<br><br>  Borrower added on 12/22/11 > I plan to use this money to finance the motorcycle i am looking at. I plan to have it paid off as soon as possible/when i sell my old bike.I only need this money because the deal im looking at is to good to pass up. I have finished college with an associates degree in business and its takingmeplaces<br></td><td>car               </td><td>bike                                 </td><td>309xx     </td><td>GA          </td><td style=\"text-align: right;\"> 1   </td><td style=\"text-align: right;\">            0</td><td>1999-04-01 00:00:00</td><td style=\"text-align: right;\">               5</td><td>                        </td><td>                        </td><td style=\"text-align: right;\">         3</td><td style=\"text-align: right;\">        0</td><td style=\"text-align: right;\">       1687</td><td>9.4%        </td><td style=\"text-align: right;\">          4</td><td>f                    </td><td style=\"text-align: right;\">       0   </td><td style=\"text-align: right;\">           0   </td><td style=\"text-align: right;\">      1008.71</td><td style=\"text-align: right;\">          1008.71</td><td style=\"text-align: right;\">           456.46</td><td style=\"text-align: right;\">         435.17</td><td style=\"text-align: right;\">                0   </td><td style=\"text-align: right;\">      117.08</td><td style=\"text-align: right;\">                     1.11</td><td>2013-04-01 00:00:00</td><td style=\"text-align: right;\">           119.66</td><td>                   </td><td>2013-09-01 00:00:00 </td><td style=\"text-align: right;\">                           0</td><td>                             </td><td style=\"text-align: right;\">            1</td></tr>\n",
       "<tr><td style=\"text-align: right;\">1.07718e+06</td><td style=\"text-align: right;\">1.31352e+06</td><td style=\"text-align: right;\">       2400</td><td style=\"text-align: right;\">         2400</td><td style=\"text-align: right;\">             2400</td><td>36 months</td><td>15.96%    </td><td style=\"text-align: right;\">        84.33</td><td>C      </td><td>C5         </td><td>                         </td><td>10+ years   </td><td>RENT            </td><td style=\"text-align: right;\">       12252</td><td>not verified            </td><td>2011-12-01 00:00:00</td><td>Fully Paid   </td><td>n           </td><td>https://www.lendingclub.com/browse/loanDetail.action?loan_id=1077175</td><td>                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                            </td><td>small_business    </td><td>real estate business                 </td><td>606xx     </td><td>IL          </td><td style=\"text-align: right;\"> 8.72</td><td style=\"text-align: right;\">            0</td><td>2001-11-01 00:00:00</td><td style=\"text-align: right;\">               2</td><td>                        </td><td>                        </td><td style=\"text-align: right;\">         2</td><td style=\"text-align: right;\">        0</td><td style=\"text-align: right;\">       2956</td><td>98.5%       </td><td style=\"text-align: right;\">         10</td><td>f                    </td><td style=\"text-align: right;\">       0   </td><td style=\"text-align: right;\">           0   </td><td style=\"text-align: right;\">      3003.65</td><td style=\"text-align: right;\">          3003.65</td><td style=\"text-align: right;\">          2400   </td><td style=\"text-align: right;\">         603.65</td><td style=\"text-align: right;\">                0   </td><td style=\"text-align: right;\">        0   </td><td style=\"text-align: right;\">                     0   </td><td>2014-06-01 00:00:00</td><td style=\"text-align: right;\">           649.91</td><td>                   </td><td>2015-06-01 00:00:00 </td><td style=\"text-align: right;\">                           0</td><td>                             </td><td style=\"text-align: right;\">            1</td></tr>\n",
       "<tr><td style=\"text-align: right;\">1.07686e+06</td><td style=\"text-align: right;\">1.27718e+06</td><td style=\"text-align: right;\">      10000</td><td style=\"text-align: right;\">        10000</td><td style=\"text-align: right;\">            10000</td><td>36 months</td><td>13.49%    </td><td style=\"text-align: right;\">       339.31</td><td>C      </td><td>C1         </td><td>AIR RESOURCES BOARD      </td><td>10+ years   </td><td>RENT            </td><td style=\"text-align: right;\">       49200</td><td>VERIFIED - income source</td><td>2011-12-01 00:00:00</td><td>Fully Paid   </td><td>n           </td><td>https://www.lendingclub.com/browse/loanDetail.action?loan_id=1076863</td><td>Borrower added on 12/21/11 > to pay for property tax (borrow from friend, need to pay back) & central A/C need to be replace. I'm very sorry to let my loan expired last time.<br>                                                                                                                                                                                                                                                                                                                                                                                                                          </td><td>other             </td><td>personel                             </td><td>917xx     </td><td>CA          </td><td style=\"text-align: right;\">20   </td><td style=\"text-align: right;\">            0</td><td>1996-02-01 00:00:00</td><td style=\"text-align: right;\">               1</td><td>35                      </td><td>                        </td><td style=\"text-align: right;\">        10</td><td style=\"text-align: right;\">        0</td><td style=\"text-align: right;\">       5598</td><td>21%         </td><td style=\"text-align: right;\">         37</td><td>f                    </td><td style=\"text-align: right;\">       0   </td><td style=\"text-align: right;\">           0   </td><td style=\"text-align: right;\">     12226.3 </td><td style=\"text-align: right;\">         12226.3 </td><td style=\"text-align: right;\">         10000   </td><td style=\"text-align: right;\">        2209.33</td><td style=\"text-align: right;\">               16.97</td><td style=\"text-align: right;\">        0   </td><td style=\"text-align: right;\">                     0   </td><td>2015-01-01 00:00:00</td><td style=\"text-align: right;\">           357.48</td><td>                   </td><td>2015-01-01 00:00:00 </td><td style=\"text-align: right;\">                           0</td><td>                             </td><td style=\"text-align: right;\">            1</td></tr>\n",
       "<tr><td style=\"text-align: right;\">1.07536e+06</td><td style=\"text-align: right;\">1.31175e+06</td><td style=\"text-align: right;\">       3000</td><td style=\"text-align: right;\">         3000</td><td style=\"text-align: right;\">             3000</td><td>60 months</td><td>12.69%    </td><td style=\"text-align: right;\">        67.79</td><td>B      </td><td>B5         </td><td>University Medical Group </td><td>1 year      </td><td>RENT            </td><td style=\"text-align: right;\">       80000</td><td>VERIFIED - income source</td><td>2011-12-01 00:00:00</td><td>Current      </td><td>n           </td><td>https://www.lendingclub.com/browse/loanDetail.action?loan_id=1075358</td><td>Borrower added on 12/21/11 > I plan on combining three large interest bills together and freeing up some extra each month to pay toward other bills.  I've always been a good payor but have found myself needing to make adjustments to my budget due to a medical scare. My job is very stable, I love it.<br>                                                                                                                                                                                                                                                                                            </td><td>other             </td><td>Personal                             </td><td>972xx     </td><td>OR          </td><td style=\"text-align: right;\">17.94</td><td style=\"text-align: right;\">            0</td><td>1996-01-01 00:00:00</td><td style=\"text-align: right;\">               0</td><td>38                      </td><td>                        </td><td style=\"text-align: right;\">        15</td><td style=\"text-align: right;\">        0</td><td style=\"text-align: right;\">      27783</td><td>53.9%       </td><td style=\"text-align: right;\">         38</td><td>f                    </td><td style=\"text-align: right;\">    1168.04</td><td style=\"text-align: right;\">        1168.04</td><td style=\"text-align: right;\">      2767.64</td><td style=\"text-align: right;\">          2767.64</td><td style=\"text-align: right;\">          1831.96</td><td style=\"text-align: right;\">         935.68</td><td style=\"text-align: right;\">                0   </td><td style=\"text-align: right;\">        0   </td><td style=\"text-align: right;\">                     0   </td><td>2015-06-01 00:00:00</td><td style=\"text-align: right;\">            67.79</td><td>2015-07-01 00:00:00</td><td>2015-06-01 00:00:00 </td><td style=\"text-align: right;\">                           0</td><td>                             </td><td style=\"text-align: right;\">            1</td></tr>\n",
       "<tr><td style=\"text-align: right;\">1.07527e+06</td><td style=\"text-align: right;\">1.31144e+06</td><td style=\"text-align: right;\">       5000</td><td style=\"text-align: right;\">         5000</td><td style=\"text-align: right;\">             5000</td><td>36 months</td><td>7.90%     </td><td style=\"text-align: right;\">       156.46</td><td>A      </td><td>A4         </td><td>Veolia Transportaton     </td><td>3 years     </td><td>RENT            </td><td style=\"text-align: right;\">       36000</td><td>VERIFIED - income source</td><td>2011-12-01 00:00:00</td><td>Fully Paid   </td><td>n           </td><td>https://www.lendingclub.com/browse/loanDetail.action?loan_id=1075269</td><td>                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                            </td><td>wedding           </td><td>My wedding loan I promise to pay back</td><td>852xx     </td><td>AZ          </td><td style=\"text-align: right;\">11.2 </td><td style=\"text-align: right;\">            0</td><td>2004-11-01 00:00:00</td><td style=\"text-align: right;\">               3</td><td>                        </td><td>                        </td><td style=\"text-align: right;\">         9</td><td style=\"text-align: right;\">        0</td><td style=\"text-align: right;\">       7963</td><td>28.3%       </td><td style=\"text-align: right;\">         12</td><td>f                    </td><td style=\"text-align: right;\">       0   </td><td style=\"text-align: right;\">           0   </td><td style=\"text-align: right;\">      5631.38</td><td style=\"text-align: right;\">          5631.38</td><td style=\"text-align: right;\">          5000   </td><td style=\"text-align: right;\">         631.38</td><td style=\"text-align: right;\">                0   </td><td style=\"text-align: right;\">        0   </td><td style=\"text-align: right;\">                     0   </td><td>2015-01-01 00:00:00</td><td style=\"text-align: right;\">           161.03</td><td>                   </td><td>2015-06-01 00:00:00 </td><td style=\"text-align: right;\">                           0</td><td>                             </td><td style=\"text-align: right;\">            1</td></tr>\n",
       "<tr><td style=\"text-align: right;\">1.06964e+06</td><td style=\"text-align: right;\">1.30474e+06</td><td style=\"text-align: right;\">       7000</td><td style=\"text-align: right;\">         7000</td><td style=\"text-align: right;\">             7000</td><td>60 months</td><td>15.96%    </td><td style=\"text-align: right;\">       170.08</td><td>C      </td><td>C5         </td><td>Southern Star Photography</td><td>8 years     </td><td>RENT            </td><td style=\"text-align: right;\">       47004</td><td>not verified            </td><td>2011-12-01 00:00:00</td><td>Current      </td><td>n           </td><td>https://www.lendingclub.com/browse/loanDetail.action?loan_id=1069639</td><td>Borrower added on 12/18/11 > I am planning on using the funds to pay off two retail credit cards with 24.99% interest rates, as well as a major bank credit card with a 18.99% rate.  I pay all my bills on time, looking for a lower combined payment and lower monthly payment.<br>                                                                                                                                                                                                                                                                                                                       </td><td>debt_consolidation</td><td>Loan                                 </td><td>280xx     </td><td>NC          </td><td style=\"text-align: right;\">23.51</td><td style=\"text-align: right;\">            0</td><td>2005-07-01 00:00:00</td><td style=\"text-align: right;\">               1</td><td>                        </td><td>                        </td><td style=\"text-align: right;\">         7</td><td style=\"text-align: right;\">        0</td><td style=\"text-align: right;\">      17726</td><td>85.6%       </td><td style=\"text-align: right;\">         11</td><td>f                    </td><td style=\"text-align: right;\">    2853.21</td><td style=\"text-align: right;\">        2853.21</td><td style=\"text-align: right;\">      6946.28</td><td style=\"text-align: right;\">          6946.28</td><td style=\"text-align: right;\">          4146.8 </td><td style=\"text-align: right;\">        2799.48</td><td style=\"text-align: right;\">                0   </td><td style=\"text-align: right;\">        0   </td><td style=\"text-align: right;\">                     0   </td><td>2015-06-01 00:00:00</td><td style=\"text-align: right;\">           170.08</td><td>2015-08-01 00:00:00</td><td>2015-06-01 00:00:00 </td><td style=\"text-align: right;\">                           0</td><td>                             </td><td style=\"text-align: right;\">            1</td></tr>\n",
       "<tr><td style=\"text-align: right;\">1.07205e+06</td><td style=\"text-align: right;\">1.28869e+06</td><td style=\"text-align: right;\">       3000</td><td style=\"text-align: right;\">         3000</td><td style=\"text-align: right;\">             3000</td><td>36 months</td><td>18.64%    </td><td style=\"text-align: right;\">       109.43</td><td>E      </td><td>E1         </td><td>MKC Accounting           </td><td>9 years     </td><td>RENT            </td><td style=\"text-align: right;\">       48000</td><td>VERIFIED - income source</td><td>2011-12-01 00:00:00</td><td>Fully Paid   </td><td>n           </td><td>https://www.lendingclub.com/browse/loanDetail.action?loan_id=1072053</td><td>Borrower added on 12/16/11 > Downpayment for a car.<br>                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                     </td><td>car               </td><td>Car Downpayment                      </td><td>900xx     </td><td>CA          </td><td style=\"text-align: right;\"> 5.35</td><td style=\"text-align: right;\">            0</td><td>2007-01-01 00:00:00</td><td style=\"text-align: right;\">               2</td><td>                        </td><td>                        </td><td style=\"text-align: right;\">         4</td><td style=\"text-align: right;\">        0</td><td style=\"text-align: right;\">       8221</td><td>87.5%       </td><td style=\"text-align: right;\">          4</td><td>f                    </td><td style=\"text-align: right;\">       0   </td><td style=\"text-align: right;\">           0   </td><td style=\"text-align: right;\">      3938.14</td><td style=\"text-align: right;\">          3938.14</td><td style=\"text-align: right;\">          3000   </td><td style=\"text-align: right;\">         938.14</td><td style=\"text-align: right;\">                0   </td><td style=\"text-align: right;\">        0   </td><td style=\"text-align: right;\">                     0   </td><td>2015-01-01 00:00:00</td><td style=\"text-align: right;\">           111.34</td><td>                   </td><td>2014-12-01 00:00:00 </td><td style=\"text-align: right;\">                           0</td><td>                             </td><td style=\"text-align: right;\">            1</td></tr>\n",
       "<tr><td style=\"text-align: right;\">1.0718e+06 </td><td style=\"text-align: right;\">1.30696e+06</td><td style=\"text-align: right;\">       5600</td><td style=\"text-align: right;\">         5600</td><td style=\"text-align: right;\">             5600</td><td>60 months</td><td>21.28%    </td><td style=\"text-align: right;\">       152.39</td><td>F      </td><td>F2         </td><td>                         </td><td>4 years     </td><td>OWN             </td><td style=\"text-align: right;\">       40000</td><td>VERIFIED - income source</td><td>2011-12-01 00:00:00</td><td>Charged Off  </td><td>n           </td><td>https://www.lendingclub.com/browse/loanDetail.action?loan_id=1071795</td><td>Borrower added on 12/21/11 > I own a small home-based judgment collection business. I have 5 years experience collecting debts. I am now going from a home office to a small office. I also plan to buy a small debt portfolio (eg. $10K for $1M of debt) <br>My score is not A+ because I own my home and have no mortgage.<br>                                                                                                                                                                                                                                                                            </td><td>small_business    </td><td>Expand Business & Buy Debt Portfolio </td><td>958xx     </td><td>CA          </td><td style=\"text-align: right;\"> 5.55</td><td style=\"text-align: right;\">            0</td><td>2004-04-01 00:00:00</td><td style=\"text-align: right;\">               2</td><td>                        </td><td>                        </td><td style=\"text-align: right;\">        11</td><td style=\"text-align: right;\">        0</td><td style=\"text-align: right;\">       5210</td><td>32.6%       </td><td style=\"text-align: right;\">         13</td><td>f                    </td><td style=\"text-align: right;\">       0   </td><td style=\"text-align: right;\">           0   </td><td style=\"text-align: right;\">       646.02</td><td style=\"text-align: right;\">           646.02</td><td style=\"text-align: right;\">           162.02</td><td style=\"text-align: right;\">         294.94</td><td style=\"text-align: right;\">                0   </td><td style=\"text-align: right;\">      189.06</td><td style=\"text-align: right;\">                     2.09</td><td>2012-04-01 00:00:00</td><td style=\"text-align: right;\">           152.39</td><td>                   </td><td>2012-08-01 00:00:00 </td><td style=\"text-align: right;\">                           0</td><td>                             </td><td style=\"text-align: right;\">            1</td></tr>\n",
       "<tr><td style=\"text-align: right;\">1.07157e+06</td><td style=\"text-align: right;\">1.30672e+06</td><td style=\"text-align: right;\">       5375</td><td style=\"text-align: right;\">         5375</td><td style=\"text-align: right;\">             5350</td><td>60 months</td><td>12.69%    </td><td style=\"text-align: right;\">       121.45</td><td>B      </td><td>B5         </td><td>Starbucks                </td><td>< 1 year    </td><td>RENT            </td><td style=\"text-align: right;\">       15000</td><td>VERIFIED - income       </td><td>2011-12-01 00:00:00</td><td>Charged Off  </td><td>n           </td><td>https://www.lendingclub.com/browse/loanDetail.action?loan_id=1071570</td><td>Borrower added on 12/16/11 > I'm trying to build up my credit history. I live with my brother and have no car payment or credit cards. I am in community college and work full time. Im going to use the money to make some repairs around the house and get some maintenance done on my car.<br><br>  Borrower added on 12/20/11 > $1000 down only $4375 to go. Thanks to everyone that invested so far, looking forward to surprising my brother with the fixes around the house.<br>                                                                                                                     </td><td>other             </td><td>Building my credit history.          </td><td>774xx     </td><td>TX          </td><td style=\"text-align: right;\">18.08</td><td style=\"text-align: right;\">            0</td><td>2004-09-01 00:00:00</td><td style=\"text-align: right;\">               0</td><td>                        </td><td>                        </td><td style=\"text-align: right;\">         2</td><td style=\"text-align: right;\">        0</td><td style=\"text-align: right;\">       9279</td><td>36.5%       </td><td style=\"text-align: right;\">          3</td><td>f                    </td><td style=\"text-align: right;\">       0   </td><td style=\"text-align: right;\">           0   </td><td style=\"text-align: right;\">      1476.19</td><td style=\"text-align: right;\">          1469.34</td><td style=\"text-align: right;\">           673.48</td><td style=\"text-align: right;\">         533.42</td><td style=\"text-align: right;\">                0   </td><td style=\"text-align: right;\">      269.29</td><td style=\"text-align: right;\">                     2.52</td><td>2012-11-01 00:00:00</td><td style=\"text-align: right;\">           121.45</td><td>                   </td><td>2013-03-01 00:00:00 </td><td style=\"text-align: right;\">                           0</td><td>                             </td><td style=\"text-align: right;\">            1</td></tr>\n",
       "</tbody>\n",
       "</table>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/plain": []
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "data.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Our dataset is still in the raw form.  We will first need to create our target variable which indicates if the loan was fully paid off.  There are, however, loans that are still currently being paid off.  We will want to remove these loans from our data since we do not know if they will be fully paid off or not in the future."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "collapsed": false,
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<table>\n",
       "<thead>\n",
       "<tr><th>loan_status                                             </th><th style=\"text-align: right;\">  Count</th></tr>\n",
       "</thead>\n",
       "<tbody>\n",
       "<tr><td>Charged Off                                             </td><td style=\"text-align: right;\">   5435</td></tr>\n",
       "<tr><td>Current                                                 </td><td style=\"text-align: right;\">   3351</td></tr>\n",
       "<tr><td>Default                                                 </td><td style=\"text-align: right;\">      7</td></tr>\n",
       "<tr><td>Does not meet the credit policy.  Status:Charged Off    </td><td style=\"text-align: right;\">    761</td></tr>\n",
       "<tr><td>Does not meet the credit policy.  Status:Current        </td><td style=\"text-align: right;\">     53</td></tr>\n",
       "<tr><td>Does not meet the credit policy.  Status:Fully Paid     </td><td style=\"text-align: right;\">   1933</td></tr>\n",
       "<tr><td>Does not meet the credit policy.  Status:In Grace Period</td><td style=\"text-align: right;\">      2</td></tr>\n",
       "<tr><td>Fully Paid                                              </td><td style=\"text-align: right;\">  30843</td></tr>\n",
       "<tr><td>In Grace Period                                         </td><td style=\"text-align: right;\">     60</td></tr>\n",
       "<tr><td>Late (16-30 days)                                       </td><td style=\"text-align: right;\">     16</td></tr>\n",
       "</tbody>\n",
       "</table>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/plain": []
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "data[\"loan_status\"].table()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "# Drop all loans that are still in progess and therefore cannot be deemed good/bad loans\n",
    "to_remove = [\"Current\", \"In Grace Period\", \"Late (16-30 days)\", \"Late (31-120 days)\"]\n",
    "data = data[data[\"loan_status\"].isin(to_remove).logical_negation(), :]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# Make target column\n",
    "data[\"bad_loan\"] = (data[\"loan_status\"] == \"Fully Paid\").logical_negation()\n",
    "data[\"bad_loan\"] = data[\"bad_loan\"].asfactor()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Now we have a binary target column called `bad_loan` indicate if the loan was not fully paid off."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<table>\n",
       "<thead>\n",
       "<tr><th style=\"text-align: right;\">  bad_loan</th><th style=\"text-align: right;\">  Count</th></tr>\n",
       "</thead>\n",
       "<tbody>\n",
       "<tr><td style=\"text-align: right;\">         0</td><td style=\"text-align: right;\">  30843</td></tr>\n",
       "<tr><td style=\"text-align: right;\">         1</td><td style=\"text-align: right;\">   8194</td></tr>\n",
       "</tbody>\n",
       "</table>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/plain": []
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "data[\"bad_loan\"].table()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We will now do some data cleaning.  The columns `revol_util` and `emp_length` are being treated as categoricals instead of numeric because the unit of measure is included.  Machine learning algorithms will treat this as a categorical and will not understand that 8 years is closer to 9 years than 1 year.  To recitify this, we will use H2O to remove the units of measure and convert the column to numeric."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<table>\n",
       "<thead>\n",
       "<tr><th>revol_util  </th><th>emp_length  </th></tr>\n",
       "</thead>\n",
       "<tbody>\n",
       "<tr><td>83.7%       </td><td>10+ years   </td></tr>\n",
       "<tr><td>9.4%        </td><td>< 1 year    </td></tr>\n",
       "<tr><td>98.5%       </td><td>10+ years   </td></tr>\n",
       "<tr><td>21%         </td><td>10+ years   </td></tr>\n",
       "<tr><td>28.3%       </td><td>3 years     </td></tr>\n",
       "<tr><td>87.5%       </td><td>9 years     </td></tr>\n",
       "<tr><td>32.6%       </td><td>4 years     </td></tr>\n",
       "<tr><td>36.5%       </td><td>< 1 year    </td></tr>\n",
       "<tr><td>20.6%       </td><td>5 years     </td></tr>\n",
       "<tr><td>67.1%       </td><td>10+ years   </td></tr>\n",
       "</tbody>\n",
       "</table>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/plain": []
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "data[[\"revol_util\", \"emp_length\"]]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<table>\n",
       "<thead>\n",
       "<tr><th style=\"text-align: right;\">  revol_util</th></tr>\n",
       "</thead>\n",
       "<tbody>\n",
       "<tr><td style=\"text-align: right;\">         943</td></tr>\n",
       "<tr><td style=\"text-align: right;\">        1012</td></tr>\n",
       "<tr><td style=\"text-align: right;\">        1105</td></tr>\n",
       "<tr><td style=\"text-align: right;\">         191</td></tr>\n",
       "<tr><td style=\"text-align: right;\">         277</td></tr>\n",
       "<tr><td style=\"text-align: right;\">         981</td></tr>\n",
       "<tr><td style=\"text-align: right;\">         336</td></tr>\n",
       "<tr><td style=\"text-align: right;\">         381</td></tr>\n",
       "<tr><td style=\"text-align: right;\">         187</td></tr>\n",
       "<tr><td style=\"text-align: right;\">         744</td></tr>\n",
       "</tbody>\n",
       "</table>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Data cleaning of revol_util\n",
    "\n",
    "data[\"revol_util\"] = data[\"revol_util\"].gsub(pattern = \"%\", replacement = \"\")\n",
    "data[\"revol_util\"] = data[\"revol_util\"].trim() \n",
    "data[\"revol_util\"] = data[\"revol_util\"].asnumeric() \n",
    "data[\"revol_util\"].show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<table>\n",
       "<thead>\n",
       "<tr><th style=\"text-align: right;\">  emp_length</th></tr>\n",
       "</thead>\n",
       "<tbody>\n",
       "<tr><td style=\"text-align: right;\">          10</td></tr>\n",
       "<tr><td style=\"text-align: right;\">           0</td></tr>\n",
       "<tr><td style=\"text-align: right;\">          10</td></tr>\n",
       "<tr><td style=\"text-align: right;\">          10</td></tr>\n",
       "<tr><td style=\"text-align: right;\">           3</td></tr>\n",
       "<tr><td style=\"text-align: right;\">           9</td></tr>\n",
       "<tr><td style=\"text-align: right;\">           4</td></tr>\n",
       "<tr><td style=\"text-align: right;\">           0</td></tr>\n",
       "<tr><td style=\"text-align: right;\">           5</td></tr>\n",
       "<tr><td style=\"text-align: right;\">          10</td></tr>\n",
       "</tbody>\n",
       "</table>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Data cleaning of emp_length\n",
    "data[\"emp_length\"] = data[\"emp_length\"].ascharacter()\n",
    "data[\"emp_length\"] = data[\"emp_length\"].gsub(pattern = \"([ ]*+[a-zA-Z].*)|(n/a)\", replacement = \"\") \n",
    "data[\"emp_length\"] = data[\"emp_length\"].trim()\n",
    "data[\"emp_length\"] = data[\"emp_length\"].gsub(pattern = \"< 1\", replacement = \"0\")\n",
    "data[\"emp_length\"] = data[\"emp_length\"].gsub(pattern = \"10\\\\+\", replacement = \"10\")\n",
    "data[\"emp_length\"] = data[\"emp_length\"].asnumeric()\n",
    "data[\"emp_length\"].show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Our last cleaning step will be to remove all columns that are not relevant to our use case.  Some columns we will not use because they have data leakage. Data leakage means the column has information that we will not know at the time a person applies for a loan."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# Filter to relevant columns\n",
    "cols = [\"purpose\", \"desc\", \"addr_state\", \"term\", \"emp_length\", \"home_ownership\", \"annual_inc\", \"dti\", \"delinq_2yrs\", \n",
    "        \"revol_util\", \"total_acc\", \"bad_loan\", \"issue_d\", \"earliest_cr_line\", \"verification_status\"]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<table>\n",
       "<thead>\n",
       "<tr><th>purpose           </th><th>desc                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                        </th><th>addr_state  </th><th>term     </th><th style=\"text-align: right;\">  emp_length</th><th>home_ownership  </th><th style=\"text-align: right;\">  annual_inc</th><th style=\"text-align: right;\">  dti</th><th style=\"text-align: right;\">  delinq_2yrs</th><th style=\"text-align: right;\">  revol_util</th><th style=\"text-align: right;\">  total_acc</th><th style=\"text-align: right;\">  bad_loan</th><th>issue_d            </th><th>earliest_cr_line   </th><th>verification_status     </th></tr>\n",
       "</thead>\n",
       "<tbody>\n",
       "<tr><td>credit_card       </td><td>Borrower added on 12/22/11 > I need to upgrade my business technologies.<br>                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                </td><td>AZ          </td><td>36 months</td><td style=\"text-align: right;\">          10</td><td>RENT            </td><td style=\"text-align: right;\">       24000</td><td style=\"text-align: right;\">27.65</td><td style=\"text-align: right;\">            0</td><td style=\"text-align: right;\">         943</td><td style=\"text-align: right;\">          9</td><td style=\"text-align: right;\">         0</td><td>2011-12-01 00:00:00</td><td>1985-01-01 00:00:00</td><td>VERIFIED - income       </td></tr>\n",
       "<tr><td>car               </td><td>Borrower added on 12/22/11 > I plan to use this money to finance the motorcycle i am looking at. I plan to have it paid off as soon as possible/when i sell my old bike. I only need this money because the deal im looking at is to good to pass up.<br><br>  Borrower added on 12/22/11 > I plan to use this money to finance the motorcycle i am looking at. I plan to have it paid off as soon as possible/when i sell my old bike.I only need this money because the deal im looking at is to good to pass up. I have finished college with an associates degree in business and its takingmeplaces<br></td><td>GA          </td><td>60 months</td><td style=\"text-align: right;\">           0</td><td>RENT            </td><td style=\"text-align: right;\">       30000</td><td style=\"text-align: right;\"> 1   </td><td style=\"text-align: right;\">            0</td><td style=\"text-align: right;\">        1012</td><td style=\"text-align: right;\">          4</td><td style=\"text-align: right;\">         1</td><td>2011-12-01 00:00:00</td><td>1999-04-01 00:00:00</td><td>VERIFIED - income source</td></tr>\n",
       "<tr><td>small_business    </td><td>                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                            </td><td>IL          </td><td>36 months</td><td style=\"text-align: right;\">          10</td><td>RENT            </td><td style=\"text-align: right;\">       12252</td><td style=\"text-align: right;\"> 8.72</td><td style=\"text-align: right;\">            0</td><td style=\"text-align: right;\">        1105</td><td style=\"text-align: right;\">         10</td><td style=\"text-align: right;\">         0</td><td>2011-12-01 00:00:00</td><td>2001-11-01 00:00:00</td><td>not verified            </td></tr>\n",
       "<tr><td>other             </td><td>Borrower added on 12/21/11 > to pay for property tax (borrow from friend, need to pay back) & central A/C need to be replace. I'm very sorry to let my loan expired last time.<br>                                                                                                                                                                                                                                                                                                                                                                                                                          </td><td>CA          </td><td>36 months</td><td style=\"text-align: right;\">          10</td><td>RENT            </td><td style=\"text-align: right;\">       49200</td><td style=\"text-align: right;\">20   </td><td style=\"text-align: right;\">            0</td><td style=\"text-align: right;\">         191</td><td style=\"text-align: right;\">         37</td><td style=\"text-align: right;\">         0</td><td>2011-12-01 00:00:00</td><td>1996-02-01 00:00:00</td><td>VERIFIED - income source</td></tr>\n",
       "<tr><td>wedding           </td><td>                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                            </td><td>AZ          </td><td>36 months</td><td style=\"text-align: right;\">           3</td><td>RENT            </td><td style=\"text-align: right;\">       36000</td><td style=\"text-align: right;\">11.2 </td><td style=\"text-align: right;\">            0</td><td style=\"text-align: right;\">         277</td><td style=\"text-align: right;\">         12</td><td style=\"text-align: right;\">         0</td><td>2011-12-01 00:00:00</td><td>2004-11-01 00:00:00</td><td>VERIFIED - income source</td></tr>\n",
       "<tr><td>car               </td><td>Borrower added on 12/16/11 > Downpayment for a car.<br>                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                     </td><td>CA          </td><td>36 months</td><td style=\"text-align: right;\">           9</td><td>RENT            </td><td style=\"text-align: right;\">       48000</td><td style=\"text-align: right;\"> 5.35</td><td style=\"text-align: right;\">            0</td><td style=\"text-align: right;\">         981</td><td style=\"text-align: right;\">          4</td><td style=\"text-align: right;\">         0</td><td>2011-12-01 00:00:00</td><td>2007-01-01 00:00:00</td><td>VERIFIED - income source</td></tr>\n",
       "<tr><td>small_business    </td><td>Borrower added on 12/21/11 > I own a small home-based judgment collection business. I have 5 years experience collecting debts. I am now going from a home office to a small office. I also plan to buy a small debt portfolio (eg. $10K for $1M of debt) <br>My score is not A+ because I own my home and have no mortgage.<br>                                                                                                                                                                                                                                                                            </td><td>CA          </td><td>60 months</td><td style=\"text-align: right;\">           4</td><td>OWN             </td><td style=\"text-align: right;\">       40000</td><td style=\"text-align: right;\"> 5.55</td><td style=\"text-align: right;\">            0</td><td style=\"text-align: right;\">         336</td><td style=\"text-align: right;\">         13</td><td style=\"text-align: right;\">         1</td><td>2011-12-01 00:00:00</td><td>2004-04-01 00:00:00</td><td>VERIFIED - income source</td></tr>\n",
       "<tr><td>other             </td><td>Borrower added on 12/16/11 > I'm trying to build up my credit history. I live with my brother and have no car payment or credit cards. I am in community college and work full time. Im going to use the money to make some repairs around the house and get some maintenance done on my car.<br><br>  Borrower added on 12/20/11 > $1000 down only $4375 to go. Thanks to everyone that invested so far, looking forward to surprising my brother with the fixes around the house.<br>                                                                                                                     </td><td>TX          </td><td>60 months</td><td style=\"text-align: right;\">           0</td><td>RENT            </td><td style=\"text-align: right;\">       15000</td><td style=\"text-align: right;\">18.08</td><td style=\"text-align: right;\">            0</td><td style=\"text-align: right;\">         381</td><td style=\"text-align: right;\">          3</td><td style=\"text-align: right;\">         1</td><td>2011-12-01 00:00:00</td><td>2004-09-01 00:00:00</td><td>VERIFIED - income       </td></tr>\n",
       "<tr><td>debt_consolidation</td><td>Borrower added on 12/15/11 > I had recived a loan from Citi Financial about a year ago, I was paying 29.99 intrest, so the refinance is to cut that rate since cleaning up my credit I have been paying everything on time as shown on my credit report<br>                                                                                                                                                                                                                                                                                                                                                 </td><td>AZ          </td><td>60 months</td><td style=\"text-align: right;\">           5</td><td>OWN             </td><td style=\"text-align: right;\">       72000</td><td style=\"text-align: right;\">16.12</td><td style=\"text-align: right;\">            0</td><td style=\"text-align: right;\">         187</td><td style=\"text-align: right;\">         23</td><td style=\"text-align: right;\">         0</td><td>2011-12-01 00:00:00</td><td>1998-01-01 00:00:00</td><td>not verified            </td></tr>\n",
       "<tr><td>debt_consolidation</td><td>                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                            </td><td>CA          </td><td>36 months</td><td style=\"text-align: right;\">          10</td><td>OWN             </td><td style=\"text-align: right;\">       75000</td><td style=\"text-align: right;\">10.78</td><td style=\"text-align: right;\">            0</td><td style=\"text-align: right;\">         744</td><td style=\"text-align: right;\">         34</td><td style=\"text-align: right;\">         0</td><td>2011-12-01 00:00:00</td><td>1989-10-01 00:00:00</td><td>VERIFIED - income source</td></tr>\n",
       "</tbody>\n",
       "</table>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/plain": []
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "data = data[cols]\n",
    "data.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Step 2. Train Baseline Model\n",
    "\n",
    "We will train an initial model.  The goal of our initial model is to provide a baseline AUC that we will try to beat with our feature engineering. \n",
    "\n",
    "We will start by splitting the data into training and testing by time.  Our testing data will consist of the last two months of data (2011/11 and 2011/12)."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "is_train = ~((data[\"issue_d\"].year() == 2011) and (data[\"issue_d\"].month() >= 11))\n",
    "\n",
    "train = data[is_train == 1]\n",
    "test = data[is_train == 0]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "target = \"bad_loan\"\n",
    "predictors = list(set(train.col_names) - set([target]))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "gbm Model Build progress: |███████████████████████████████████████████████| 100%\n"
     ]
    }
   ],
   "source": [
    "# Train initial GBM model\n",
    "from h2o.estimators import H2OGradientBoostingEstimator\n",
    "\n",
    "gbm_baseline = H2OGradientBoostingEstimator(seed = 1234, model_id = \"gbm_baseline.hex\", \n",
    "                                            ntrees = 500, \n",
    "                                            stopping_rounds = 10, score_tree_interval = 5, # early stopping\n",
    "                                            stopping_metric = \"AUC\",\n",
    "                                            nfolds = 5 # 5-fold cross validation\n",
    "                                           )\n",
    "gbm_baseline.train(x = predictors, y = target, training_frame = train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "AUC on CV: 0.719\n"
     ]
    }
   ],
   "source": [
    "print(\"AUC on CV: \" + str(round(gbm_baseline.auc(xval = True), 3)))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAA3EAAAJTCAYAAACmZ5/FAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n7D64VEj6k6wAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x11694a0f0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "gbm_baseline.varimp_plot()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Step 3. Convert Text to Word Embeddings\n",
    "\n",
    "Our dataset has a text column called `desc`.  We will convert this string to word embeddings so that it can be more useful to our supervised learning model."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {
    "collapsed": false,
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<table>\n",
       "<thead>\n",
       "<tr><th>desc                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                        </th></tr>\n",
       "</thead>\n",
       "<tbody>\n",
       "<tr><td>Borrower added on 12/22/11 > I need to upgrade my business technologies.<br>                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                </td></tr>\n",
       "<tr><td>Borrower added on 12/22/11 > I plan to use this money to finance the motorcycle i am looking at. I plan to have it paid off as soon as possible/when i sell my old bike. I only need this money because the deal im looking at is to good to pass up.<br><br>  Borrower added on 12/22/11 > I plan to use this money to finance the motorcycle i am looking at. I plan to have it paid off as soon as possible/when i sell my old bike.I only need this money because the deal im looking at is to good to pass up. I have finished college with an associates degree in business and its takingmeplaces<br></td></tr>\n",
       "<tr><td>                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                            </td></tr>\n",
       "<tr><td>Borrower added on 12/21/11 > to pay for property tax (borrow from friend, need to pay back) & central A/C need to be replace. I'm very sorry to let my loan expired last time.<br>                                                                                                                                                                                                                                                                                                                                                                                                                          </td></tr>\n",
       "<tr><td>                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                            </td></tr>\n",
       "<tr><td>Borrower added on 12/16/11 > Downpayment for a car.<br>                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                     </td></tr>\n",
       "<tr><td>Borrower added on 12/21/11 > I own a small home-based judgment collection business. I have 5 years experience collecting debts. I am now going from a home office to a small office. I also plan to buy a small debt portfolio (eg. $10K for $1M of debt) <br>My score is not A+ because I own my home and have no mortgage.<br>                                                                                                                                                                                                                                                                            </td></tr>\n",
       "<tr><td>Borrower added on 12/16/11 > I'm trying to build up my credit history. I live with my brother and have no car payment or credit cards. I am in community college and work full time. Im going to use the money to make some repairs around the house and get some maintenance done on my car.<br><br>  Borrower added on 12/20/11 > $1000 down only $4375 to go. Thanks to everyone that invested so far, looking forward to surprising my brother with the fixes around the house.<br>                                                                                                                     </td></tr>\n",
       "<tr><td>Borrower added on 12/15/11 > I had recived a loan from Citi Financial about a year ago, I was paying 29.99 intrest, so the refinance is to cut that rate since cleaning up my credit I have been paying everything on time as shown on my credit report<br>                                                                                                                                                                                                                                                                                                                                                 </td></tr>\n",
       "<tr><td>                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                            </td></tr>\n",
       "</tbody>\n",
       "</table>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/plain": []
     },
     "execution_count": 18,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "data[\"desc\"]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We will first examine the performance of our model if we remove the `desc` variable from the model.  "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "gbm Model Build progress: |███████████████████████████████████████████████| 100%\n"
     ]
    }
   ],
   "source": [
    "# Train initial GBM model\n",
    "gbm_no_desc = H2OGradientBoostingEstimator(seed = 1234, model_id = \"gbm_baseline.hex\", \n",
    "                                           ntrees = 500, \n",
    "                                           stopping_rounds = 10, score_tree_interval = 5, # early stopping\n",
    "                                           stopping_metric = \"AUC\",\n",
    "                                           nfolds = 5 # 5-fold cross validation\n",
    "                                          )\n",
    "gbm_no_desc.train(x = list(set(predictors) - set([\"desc\"])), y = target, training_frame = train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "AUC on CV - Baseline: 0.719\n",
      "AUC on CV - without Desc: 0.721\n"
     ]
    }
   ],
   "source": [
    "print(\"AUC on CV - Baseline: \" + str(round(gbm_baseline.auc(xval = True), 3)))\n",
    "print(\"AUC on CV - without Desc: \" + str(round(gbm_no_desc.auc(xval = True), 3)))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "The performance of our model is much better if we do not use `desc` as a predictor.  This could be because the algorithm has a hard time understanding the text. We will instead convert it to numeric vectors by using the Word2Vec algorithm.\n",
    "\n",
    "Our first step will be to tokenize the `desc` column.  This splits the descriptions by word.  Our function will also:\n",
    "\n",
    "* convert all words to lowercase\n",
    "* remove words with numbers\n",
    "* remove words with only 1 letter\n",
    "* remove stop words (stop words are defined by the nltk package)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['i', 'me', 'my', 'myself', 'we']"
      ]
     },
     "execution_count": 21,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from nltk.corpus import stopwords\n",
    "STOP_WORDS = stopwords.words('english')\n",
    "STOP_WORDS[0:5]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def tokenize(sentences, stop_word = STOP_WORDS):\n",
    "    tokenized = sentences.tokenize(\"\\\\W+\")\n",
    "    tokenized_lower = tokenized.tolower()\n",
    "    tokenized_filtered = tokenized_lower[(tokenized_lower.nchar() >= 2) | (tokenized_lower.isna()),:]\n",
    "    tokenized_words = tokenized_filtered[tokenized_filtered.grep(\"[0-9]\",invert=True,output_logical=True),:]\n",
    "    tokenized_words = tokenized_words[(tokenized_words.isna()) | (~ tokenized_words.isin(STOP_WORDS)),:]\n",
    "    return tokenized_words"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "# Break desc into sequence of words\n",
    "desc = tokenize(train[\"desc\"].ascharacter())"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Now we have our the tokenized `desc`, we can train our Word2Vec model.  This will learn the word embeddings that will represent each word."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "word2vec Model Build progress: |██████████████████████████████████████████| 100%\n"
     ]
    }
   ],
   "source": [
    "from h2o.estimators.word2vec import H2OWord2vecEstimator\n",
    "\n",
    "# This takes time to run - left commented out\n",
    "w2v_model = H2OWord2vecEstimator(vec_size = 100, model_id = \"w2v.hex\")\n",
    "w2v_model.train(training_frame=desc)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We can use the Word2Vec model to find synonyms.  Synonyms are defined as words with small cosine distance between their word embeddings.  Words that have similar word embeddings should have similar meanings.  \n",
    "\n",
    "We can test the understanding of our Word2Vec model by finding synonyms for the words `wedding` and `car`."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "OrderedDict([('honeymoon', 0.8067679405212402),\n",
       "             ('ceremony', 0.7539779543876648),\n",
       "             ('guests', 0.6756937503814697),\n",
       "             ('invitations', 0.6737778186798096),\n",
       "             ('reception', 0.6730184555053711)])"
      ]
     },
     "execution_count": 25,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Sanity check - find synonyms for the word 'wedding'\n",
    "w2v_model.find_synonyms(\"wedding\", count = 5)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "OrderedDict([('auto', 0.7344753742218018),\n",
       "             ('vehicle', 0.7109878659248352),\n",
       "             ('convertible', 0.6764197945594788),\n",
       "             ('lexus', 0.6660585999488831),\n",
       "             ('mercedes', 0.6648411154747009)])"
      ]
     },
     "execution_count": 26,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Sanity check - find synonyms for the word 'car'\n",
    "w2v_model.find_synonyms(\"car\", count = 5)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Now we can add the word embeddings as new features in our data.  There are multiple words for each description so we will need to aggregate our word embeddings so that there is a single vector of word embeddings per loan.  We will do this by averaging the word embeddings using the `aggregate_method = \"AVERAGE\"`."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# Calculate a vector for each desc\n",
    "train_desc_vecs = w2v_model.transform(desc, aggregate_method = \"AVERAGE\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<table>\n",
       "<thead>\n",
       "<tr><th style=\"text-align: right;\">         C1</th><th style=\"text-align: right;\">         C2</th><th style=\"text-align: right;\">          C3</th><th style=\"text-align: right;\">          C4</th><th style=\"text-align: right;\">          C5</th><th style=\"text-align: right;\">          C6</th><th style=\"text-align: right;\">        C7</th><th style=\"text-align: right;\">         C8</th><th style=\"text-align: right;\">          C9</th><th style=\"text-align: right;\">         C10</th><th style=\"text-align: right;\">        C11</th><th style=\"text-align: right;\">        C12</th><th style=\"text-align: right;\">        C13</th><th style=\"text-align: right;\">        C14</th><th style=\"text-align: right;\">          C15</th><th style=\"text-align: right;\">        C16</th><th style=\"text-align: right;\">         C17</th><th style=\"text-align: right;\">         C18</th><th style=\"text-align: right;\">        C19</th><th style=\"text-align: right;\">        C20</th><th style=\"text-align: right;\">       C21</th><th style=\"text-align: right;\">        C22</th><th style=\"text-align: right;\">        C23</th><th style=\"text-align: right;\">         C24</th><th style=\"text-align: right;\">        C25</th><th style=\"text-align: right;\">          C26</th><th style=\"text-align: right;\">         C27</th><th style=\"text-align: right;\">         C28</th><th style=\"text-align: right;\">        C29</th><th style=\"text-align: right;\">       C30</th><th style=\"text-align: right;\">         C31</th><th style=\"text-align: right;\">         C32</th><th style=\"text-align: right;\">        C33</th><th style=\"text-align: right;\">        C34</th><th style=\"text-align: right;\">         C35</th><th style=\"text-align: right;\">        C36</th><th style=\"text-align: right;\">        C37</th><th style=\"text-align: right;\">        C38</th><th style=\"text-align: right;\">        C39</th><th style=\"text-align: right;\">        C40</th><th style=\"text-align: right;\">        C41</th><th style=\"text-align: right;\">        C42</th><th style=\"text-align: right;\">       C43</th><th style=\"text-align: right;\">          C44</th><th style=\"text-align: right;\">        C45</th><th style=\"text-align: right;\">        C46</th><th style=\"text-align: right;\">       C47</th><th style=\"text-align: right;\">        C48</th><th style=\"text-align: right;\">         C49</th><th style=\"text-align: right;\">       C50</th><th style=\"text-align: right;\">       C51</th><th style=\"text-align: right;\">         C52</th><th style=\"text-align: right;\">        C53</th><th style=\"text-align: right;\">        C54</th><th style=\"text-align: right;\">       C55</th><th style=\"text-align: right;\">       C56</th><th style=\"text-align: right;\">        C57</th><th style=\"text-align: right;\">          C58</th><th style=\"text-align: right;\">         C59</th><th style=\"text-align: right;\">        C60</th><th style=\"text-align: right;\">       C61</th><th style=\"text-align: right;\">        C62</th><th style=\"text-align: right;\">        C63</th><th style=\"text-align: right;\">       C64</th><th style=\"text-align: right;\">        C65</th><th style=\"text-align: right;\">       C66</th><th style=\"text-align: right;\">        C67</th><th style=\"text-align: right;\">        C68</th><th style=\"text-align: right;\">        C69</th><th style=\"text-align: right;\">         C70</th><th style=\"text-align: right;\">        C71</th><th style=\"text-align: right;\">        C72</th><th style=\"text-align: right;\">          C73</th><th style=\"text-align: right;\">        C74</th><th style=\"text-align: right;\">        C75</th><th style=\"text-align: right;\">        C76</th><th style=\"text-align: right;\">        C77</th><th style=\"text-align: right;\">       C78</th><th style=\"text-align: right;\">         C79</th><th style=\"text-align: right;\">        C80</th><th style=\"text-align: right;\">         C81</th><th style=\"text-align: right;\">       C82</th><th style=\"text-align: right;\">         C83</th><th style=\"text-align: right;\">         C84</th><th style=\"text-align: right;\">        C85</th><th style=\"text-align: right;\">         C86</th><th style=\"text-align: right;\">       C87</th><th style=\"text-align: right;\">        C88</th><th style=\"text-align: right;\">        C89</th><th style=\"text-align: right;\">         C90</th><th style=\"text-align: right;\">          C91</th><th style=\"text-align: right;\">         C92</th><th style=\"text-align: right;\">        C93</th><th style=\"text-align: right;\">        C94</th><th style=\"text-align: right;\">         C95</th><th style=\"text-align: right;\">         C96</th><th style=\"text-align: right;\">         C97</th><th style=\"text-align: right;\">          C98</th><th style=\"text-align: right;\">        C99</th><th style=\"text-align: right;\">        C100</th></tr>\n",
       "</thead>\n",
       "<tbody>\n",
       "<tr><td style=\"text-align: right;\"> -0.0113489</td><td style=\"text-align: right;\"> -0.165544 </td><td style=\"text-align: right;\">  0.192567  </td><td style=\"text-align: right;\">  0.00952524</td><td style=\"text-align: right;\"> -0.0363902 </td><td style=\"text-align: right;\"> -0.00946942</td><td style=\"text-align: right;\"> -0.141947</td><td style=\"text-align: right;\">  0.27071  </td><td style=\"text-align: right;\">  0.00273591</td><td style=\"text-align: right;\"> -0.0055383 </td><td style=\"text-align: right;\">  0.0382951</td><td style=\"text-align: right;\"> -0.101176 </td><td style=\"text-align: right;\"> -0.0509894</td><td style=\"text-align: right;\">  0.0368103</td><td style=\"text-align: right;\"> -0.0803125  </td><td style=\"text-align: right;\"> -0.124853 </td><td style=\"text-align: right;\"> -0.156763  </td><td style=\"text-align: right;\"> -0.0865538 </td><td style=\"text-align: right;\"> -0.0527207</td><td style=\"text-align: right;\"> -0.192072 </td><td style=\"text-align: right;\">  0.201117</td><td style=\"text-align: right;\"> -0.0104728</td><td style=\"text-align: right;\">  0.120298 </td><td style=\"text-align: right;\">  0.0402803 </td><td style=\"text-align: right;\">  0.185125 </td><td style=\"text-align: right;\">  0.0140905  </td><td style=\"text-align: right;\">  0.0776275 </td><td style=\"text-align: right;\">  0.0214181 </td><td style=\"text-align: right;\">  0.135681 </td><td style=\"text-align: right;\">  0.201137</td><td style=\"text-align: right;\">  0.0294702 </td><td style=\"text-align: right;\">  0.0360406 </td><td style=\"text-align: right;\"> -0.0255211</td><td style=\"text-align: right;\">  0.0591759</td><td style=\"text-align: right;\"> -0.011905  </td><td style=\"text-align: right;\">  0.0260656</td><td style=\"text-align: right;\">  0.0855466</td><td style=\"text-align: right;\"> -0.0539493</td><td style=\"text-align: right;\"> -0.247695 </td><td style=\"text-align: right;\">  0.15562  </td><td style=\"text-align: right;\"> -0.226543 </td><td style=\"text-align: right;\"> -0.135879 </td><td style=\"text-align: right;\"> -0.26716 </td><td style=\"text-align: right;\">  0.047003   </td><td style=\"text-align: right;\"> -0.127963 </td><td style=\"text-align: right;\">  0.100601 </td><td style=\"text-align: right;\"> -0.17288 </td><td style=\"text-align: right;\"> -0.1064   </td><td style=\"text-align: right;\"> -0.0518995 </td><td style=\"text-align: right;\"> -0.209752</td><td style=\"text-align: right;\">  0.237233</td><td style=\"text-align: right;\"> -0.0326637 </td><td style=\"text-align: right;\">  0.160465 </td><td style=\"text-align: right;\">  0.0598639</td><td style=\"text-align: right;\"> -0.17013 </td><td style=\"text-align: right;\">  0.201529</td><td style=\"text-align: right;\">  0.0392627</td><td style=\"text-align: right;\"> -0.112598   </td><td style=\"text-align: right;\"> -0.107654  </td><td style=\"text-align: right;\"> -0.22583  </td><td style=\"text-align: right;\"> -0.17319 </td><td style=\"text-align: right;\"> -0.0394094</td><td style=\"text-align: right;\">  0.204992 </td><td style=\"text-align: right;\">  0.331102</td><td style=\"text-align: right;\"> -0.10112  </td><td style=\"text-align: right;\">  0.177852</td><td style=\"text-align: right;\"> -0.0699644</td><td style=\"text-align: right;\"> -0.165633 </td><td style=\"text-align: right;\"> -0.127031 </td><td style=\"text-align: right;\">  0.0108492 </td><td style=\"text-align: right;\"> -0.0730526</td><td style=\"text-align: right;\">  0.146502 </td><td style=\"text-align: right;\">  0.0158396  </td><td style=\"text-align: right;\">  0.0125977</td><td style=\"text-align: right;\"> -0.0721209</td><td style=\"text-align: right;\">  0.280606 </td><td style=\"text-align: right;\"> -0.146997 </td><td style=\"text-align: right;\">  0.268353</td><td style=\"text-align: right;\">  0.0154769 </td><td style=\"text-align: right;\"> -0.127565 </td><td style=\"text-align: right;\">  0.00248352</td><td style=\"text-align: right;\"> -0.147604</td><td style=\"text-align: right;\"> -0.0936263 </td><td style=\"text-align: right;\">  0.00164844</td><td style=\"text-align: right;\"> -0.0709274</td><td style=\"text-align: right;\"> -0.0170988 </td><td style=\"text-align: right;\"> -0.170003</td><td style=\"text-align: right;\"> -0.0668037</td><td style=\"text-align: right;\"> -0.262802 </td><td style=\"text-align: right;\">  0.0286525 </td><td style=\"text-align: right;\"> -0.0373488  </td><td style=\"text-align: right;\"> -0.042586  </td><td style=\"text-align: right;\"> -0.0630694</td><td style=\"text-align: right;\">  0.115427 </td><td style=\"text-align: right;\">  0.00537603</td><td style=\"text-align: right;\">  0.173609  </td><td style=\"text-align: right;\">  0.0308038 </td><td style=\"text-align: right;\"> -0.000362503</td><td style=\"text-align: right;\">  0.0642299</td><td style=\"text-align: right;\">  0.0600495 </td></tr>\n",
       "<tr><td style=\"text-align: right;\"> -0.0680932</td><td style=\"text-align: right;\"> -0.0759872</td><td style=\"text-align: right;\"> -0.0114821 </td><td style=\"text-align: right;\"> -0.0580209 </td><td style=\"text-align: right;\"> -0.00158668</td><td style=\"text-align: right;\"> -0.0991261 </td><td style=\"text-align: right;\"> -0.259689</td><td style=\"text-align: right;\">  0.120048 </td><td style=\"text-align: right;\">  0.015632  </td><td style=\"text-align: right;\"> -0.00282528</td><td style=\"text-align: right;\">  0.0378028</td><td style=\"text-align: right;\"> -0.0922352</td><td style=\"text-align: right;\">  0.0531951</td><td style=\"text-align: right;\"> -0.15258  </td><td style=\"text-align: right;\">  0.0698072  </td><td style=\"text-align: right;\"> -0.040879 </td><td style=\"text-align: right;\"> -0.0584266 </td><td style=\"text-align: right;\">  0.0751576 </td><td style=\"text-align: right;\">  0.0590438</td><td style=\"text-align: right;\"> -0.122286 </td><td style=\"text-align: right;\">  0.19953 </td><td style=\"text-align: right;\"> -0.0898973</td><td style=\"text-align: right;\"> -0.0235978</td><td style=\"text-align: right;\">  0.128029  </td><td style=\"text-align: right;\">  0.150445 </td><td style=\"text-align: right;\"> -0.0181241  </td><td style=\"text-align: right;\">  0.0599836 </td><td style=\"text-align: right;\">  0.0186632 </td><td style=\"text-align: right;\">  0.0318125</td><td style=\"text-align: right;\">  0.173153</td><td style=\"text-align: right;\">  0.00404759</td><td style=\"text-align: right;\">  0.190132  </td><td style=\"text-align: right;\">  0.0388098</td><td style=\"text-align: right;\">  0.131996 </td><td style=\"text-align: right;\"> -0.103475  </td><td style=\"text-align: right;\"> -0.143872 </td><td style=\"text-align: right;\">  0.0358567</td><td style=\"text-align: right;\">  0.0990984</td><td style=\"text-align: right;\"> -0.0816364</td><td style=\"text-align: right;\">  0.207441 </td><td style=\"text-align: right;\"> -0.0842987</td><td style=\"text-align: right;\">  0.0138828</td><td style=\"text-align: right;\"> -0.187162</td><td style=\"text-align: right;\"> -0.0723731  </td><td style=\"text-align: right;\"> -0.027074 </td><td style=\"text-align: right;\">  0.0384083</td><td style=\"text-align: right;\"> -0.259057</td><td style=\"text-align: right;\"> -0.0997713</td><td style=\"text-align: right;\">  0.0166725 </td><td style=\"text-align: right;\"> -0.165892</td><td style=\"text-align: right;\">  0.191343</td><td style=\"text-align: right;\">  0.0283441 </td><td style=\"text-align: right;\">  0.0752334</td><td style=\"text-align: right;\">  0.151495 </td><td style=\"text-align: right;\"> -0.233814</td><td style=\"text-align: right;\">  0.25716 </td><td style=\"text-align: right;\"> -0.0465797</td><td style=\"text-align: right;\"> -0.000499423</td><td style=\"text-align: right;\"> -0.105473  </td><td style=\"text-align: right;\"> -0.0420092</td><td style=\"text-align: right;\"> -0.170572</td><td style=\"text-align: right;\"> -0.0299186</td><td style=\"text-align: right;\">  0.0291544</td><td style=\"text-align: right;\">  0.162087</td><td style=\"text-align: right;\"> -0.145899 </td><td style=\"text-align: right;\">  0.210332</td><td style=\"text-align: right;\"> -0.168268 </td><td style=\"text-align: right;\"> -0.0756826</td><td style=\"text-align: right;\"> -0.0307639</td><td style=\"text-align: right;\"> -0.15868   </td><td style=\"text-align: right;\"> -0.0869254</td><td style=\"text-align: right;\">  0.0417174</td><td style=\"text-align: right;\">  0.000920731</td><td style=\"text-align: right;\"> -0.106492 </td><td style=\"text-align: right;\"> -0.127832 </td><td style=\"text-align: right;\">  0.179862 </td><td style=\"text-align: right;\"> -0.106164 </td><td style=\"text-align: right;\">  0.233166</td><td style=\"text-align: right;\">  0.0264159 </td><td style=\"text-align: right;\"> -0.0317022</td><td style=\"text-align: right;\"> -0.0917652 </td><td style=\"text-align: right;\"> -0.109464</td><td style=\"text-align: right;\"> -0.00237293</td><td style=\"text-align: right;\"> -0.0487583 </td><td style=\"text-align: right;\"> -0.151673 </td><td style=\"text-align: right;\"> -0.0403641 </td><td style=\"text-align: right;\"> -0.133419</td><td style=\"text-align: right;\"> -0.0461863</td><td style=\"text-align: right;\"> -0.0519414</td><td style=\"text-align: right;\"> -0.171982  </td><td style=\"text-align: right;\">  0.190857   </td><td style=\"text-align: right;\">  0.233163  </td><td style=\"text-align: right;\">  0.0216695</td><td style=\"text-align: right;\">  0.204684 </td><td style=\"text-align: right;\">  0.0840305 </td><td style=\"text-align: right;\">  0.0605782 </td><td style=\"text-align: right;\">  0.0472977 </td><td style=\"text-align: right;\">  0.204139   </td><td style=\"text-align: right;\">  0.104189 </td><td style=\"text-align: right;\">  0.0978648 </td></tr>\n",
       "<tr><td style=\"text-align: right;\">  0.0961154</td><td style=\"text-align: right;\"> -0.102961 </td><td style=\"text-align: right;\">  0.168668  </td><td style=\"text-align: right;\"> -0.0252266 </td><td style=\"text-align: right;\"> -0.013532  </td><td style=\"text-align: right;\"> -0.00339303</td><td style=\"text-align: right;\"> -0.037358</td><td style=\"text-align: right;\">  0.191561 </td><td style=\"text-align: right;\">  0.0440751 </td><td style=\"text-align: right;\">  0.0577892 </td><td style=\"text-align: right;\">  0.105894 </td><td style=\"text-align: right;\"> -0.0825778</td><td style=\"text-align: right;\"> -0.0363949</td><td style=\"text-align: right;\">  0.0142553</td><td style=\"text-align: right;\">  0.000500772</td><td style=\"text-align: right;\"> -0.115063 </td><td style=\"text-align: right;\"> -0.0390304 </td><td style=\"text-align: right;\"> -0.0605942 </td><td style=\"text-align: right;\"> -0.0701188</td><td style=\"text-align: right;\"> -0.151891 </td><td style=\"text-align: right;\">  0.095954</td><td style=\"text-align: right;\">  0.0353282</td><td style=\"text-align: right;\">  0.112966 </td><td style=\"text-align: right;\">  0.0464437 </td><td style=\"text-align: right;\">  0.136957 </td><td style=\"text-align: right;\">  0.0765932  </td><td style=\"text-align: right;\">  0.0927007 </td><td style=\"text-align: right;\"> -0.0421562 </td><td style=\"text-align: right;\">  0.0755096</td><td style=\"text-align: right;\">  0.265262</td><td style=\"text-align: right;\"> -0.0296224 </td><td style=\"text-align: right;\">  0.0795711 </td><td style=\"text-align: right;\"> -0.100199 </td><td style=\"text-align: right;\">  0.10125  </td><td style=\"text-align: right;\"> -0.0377494 </td><td style=\"text-align: right;\"> -0.0408401</td><td style=\"text-align: right;\">  0.084573 </td><td style=\"text-align: right;\">  0.0659957</td><td style=\"text-align: right;\"> -0.172045 </td><td style=\"text-align: right;\">  0.107924 </td><td style=\"text-align: right;\"> -0.101568 </td><td style=\"text-align: right;\"> -0.0502814</td><td style=\"text-align: right;\"> -0.300108</td><td style=\"text-align: right;\">  0.000433778</td><td style=\"text-align: right;\"> -0.0479112</td><td style=\"text-align: right;\">  0.0998032</td><td style=\"text-align: right;\"> -0.231693</td><td style=\"text-align: right;\"> -0.0691336</td><td style=\"text-align: right;\"> -0.00788841</td><td style=\"text-align: right;\"> -0.133637</td><td style=\"text-align: right;\">  0.14616 </td><td style=\"text-align: right;\"> -0.0936803 </td><td style=\"text-align: right;\">  0.12455  </td><td style=\"text-align: right;\">  0.0268465</td><td style=\"text-align: right;\"> -0.217111</td><td style=\"text-align: right;\">  0.151429</td><td style=\"text-align: right;\">  0.0657126</td><td style=\"text-align: right;\"> -0.141238   </td><td style=\"text-align: right;\"> -0.108773  </td><td style=\"text-align: right;\"> -0.293511 </td><td style=\"text-align: right;\"> -0.164663</td><td style=\"text-align: right;\">  0.0791512</td><td style=\"text-align: right;\">  0.249266 </td><td style=\"text-align: right;\">  0.234641</td><td style=\"text-align: right;\"> -0.0994324</td><td style=\"text-align: right;\">  0.137099</td><td style=\"text-align: right;\"> -0.0443866</td><td style=\"text-align: right;\"> -0.139066 </td><td style=\"text-align: right;\"> -0.0836159</td><td style=\"text-align: right;\"> -0.00190028</td><td style=\"text-align: right;\"> -0.0968992</td><td style=\"text-align: right;\">  0.173375 </td><td style=\"text-align: right;\">  0.0520609  </td><td style=\"text-align: right;\">  0.0245959</td><td style=\"text-align: right;\"> -0.0581554</td><td style=\"text-align: right;\">  0.225023 </td><td style=\"text-align: right;\"> -0.0708798</td><td style=\"text-align: right;\">  0.243419</td><td style=\"text-align: right;\"> -0.0521153 </td><td style=\"text-align: right;\"> -0.145181 </td><td style=\"text-align: right;\">  0.0191964 </td><td style=\"text-align: right;\"> -0.15413 </td><td style=\"text-align: right;\"> -0.142543  </td><td style=\"text-align: right;\"> -0.0148866 </td><td style=\"text-align: right;\"> -0.0418936</td><td style=\"text-align: right;\">  0.00338959</td><td style=\"text-align: right;\"> -0.164119</td><td style=\"text-align: right;\">  0.0136206</td><td style=\"text-align: right;\"> -0.152459 </td><td style=\"text-align: right;\"> -0.0173796 </td><td style=\"text-align: right;\">  0.000136223</td><td style=\"text-align: right;\">  0.0135341 </td><td style=\"text-align: right;\"> -0.040621 </td><td style=\"text-align: right;\">  0.16177  </td><td style=\"text-align: right;\">  0.101332  </td><td style=\"text-align: right;\">  0.152892  </td><td style=\"text-align: right;\">  0.0656017 </td><td style=\"text-align: right;\">  0.053976   </td><td style=\"text-align: right;\">  0.0247348</td><td style=\"text-align: right;\">  0.00359038</td></tr>\n",
       "<tr><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan         </td><td style=\"text-align: right;\">nan         </td><td style=\"text-align: right;\">nan         </td><td style=\"text-align: right;\">nan         </td><td style=\"text-align: right;\">nan       </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan         </td><td style=\"text-align: right;\">nan         </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan          </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan         </td><td style=\"text-align: right;\">nan         </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan       </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan         </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan          </td><td style=\"text-align: right;\">nan         </td><td style=\"text-align: right;\">nan         </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan       </td><td style=\"text-align: right;\">nan         </td><td style=\"text-align: right;\">nan         </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan         </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan       </td><td style=\"text-align: right;\">nan          </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan       </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan         </td><td style=\"text-align: right;\">nan       </td><td style=\"text-align: right;\">nan       </td><td style=\"text-align: right;\">nan         </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan       </td><td style=\"text-align: right;\">nan       </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan          </td><td style=\"text-align: right;\">nan         </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan       </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan       </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan       </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan         </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan          </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan       </td><td style=\"text-align: right;\">nan         </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan         </td><td style=\"text-align: right;\">nan       </td><td style=\"text-align: right;\">nan         </td><td style=\"text-align: right;\">nan         </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan         </td><td style=\"text-align: right;\">nan       </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan         </td><td style=\"text-align: right;\">nan          </td><td style=\"text-align: right;\">nan         </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan         </td><td style=\"text-align: right;\">nan         </td><td style=\"text-align: right;\">nan         </td><td style=\"text-align: right;\">nan          </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan         </td></tr>\n",
       "<tr><td style=\"text-align: right;\"> -0.0313728</td><td style=\"text-align: right;\"> -0.106615 </td><td style=\"text-align: right;\">  0.052357  </td><td style=\"text-align: right;\">  0.096665  </td><td style=\"text-align: right;\"> -0.0857551 </td><td style=\"text-align: right;\"> -0.0557677 </td><td style=\"text-align: right;\"> -0.146217</td><td style=\"text-align: right;\">  0.110614 </td><td style=\"text-align: right;\"> -0.0271008 </td><td style=\"text-align: right;\"> -0.0183925 </td><td style=\"text-align: right;\">  0.0777338</td><td style=\"text-align: right;\"> -0.0526362</td><td style=\"text-align: right;\"> -0.0024787</td><td style=\"text-align: right;\"> -0.0152527</td><td style=\"text-align: right;\"> -0.101276   </td><td style=\"text-align: right;\"> -0.175977 </td><td style=\"text-align: right;\"> -0.0383083 </td><td style=\"text-align: right;\"> -0.0541195 </td><td style=\"text-align: right;\"> -0.0040879</td><td style=\"text-align: right;\"> -0.124336 </td><td style=\"text-align: right;\">  0.139461</td><td style=\"text-align: right;\">  0.020566 </td><td style=\"text-align: right;\">  0.197502 </td><td style=\"text-align: right;\"> -0.00599461</td><td style=\"text-align: right;\">  0.132703 </td><td style=\"text-align: right;\">  0.173817   </td><td style=\"text-align: right;\">  0.104673  </td><td style=\"text-align: right;\"> -0.0694552 </td><td style=\"text-align: right;\">  0.0198981</td><td style=\"text-align: right;\">  0.173298</td><td style=\"text-align: right;\">  0.0639467 </td><td style=\"text-align: right;\"> -0.00030885</td><td style=\"text-align: right;\"> -0.0347486</td><td style=\"text-align: right;\">  0.132726 </td><td style=\"text-align: right;\">  0.00167256</td><td style=\"text-align: right;\"> -0.0262193</td><td style=\"text-align: right;\">  0.107838 </td><td style=\"text-align: right;\">  0.0867076</td><td style=\"text-align: right;\"> -0.162181 </td><td style=\"text-align: right;\"> -0.0330601</td><td style=\"text-align: right;\"> -0.0748626</td><td style=\"text-align: right;\"> -0.068292 </td><td style=\"text-align: right;\"> -0.125952</td><td style=\"text-align: right;\">  0.0265798  </td><td style=\"text-align: right;\"> -0.107208 </td><td style=\"text-align: right;\"> -0.0589757</td><td style=\"text-align: right;\"> -0.183891</td><td style=\"text-align: right;\"> -0.0290731</td><td style=\"text-align: right;\">  0.0298937 </td><td style=\"text-align: right;\"> -0.172724</td><td style=\"text-align: right;\">  0.216033</td><td style=\"text-align: right;\">  0.00705794</td><td style=\"text-align: right;\">  0.210061 </td><td style=\"text-align: right;\">  0.151792 </td><td style=\"text-align: right;\"> -0.207563</td><td style=\"text-align: right;\">  0.145187</td><td style=\"text-align: right;\">  0.0381566</td><td style=\"text-align: right;\"> -0.0723549  </td><td style=\"text-align: right;\"> -0.152836  </td><td style=\"text-align: right;\"> -0.217406 </td><td style=\"text-align: right;\"> -0.127132</td><td style=\"text-align: right;\">  0.0554385</td><td style=\"text-align: right;\">  0.142666 </td><td style=\"text-align: right;\">  0.224817</td><td style=\"text-align: right;\"> -0.076511 </td><td style=\"text-align: right;\">  0.191099</td><td style=\"text-align: right;\"> -0.120547 </td><td style=\"text-align: right;\"> -0.047218 </td><td style=\"text-align: right;\"> -0.0231386</td><td style=\"text-align: right;\"> -0.0241571 </td><td style=\"text-align: right;\"> -0.0739734</td><td style=\"text-align: right;\">  0.0259118</td><td style=\"text-align: right;\">  0.0239804  </td><td style=\"text-align: right;\">  0.0131198</td><td style=\"text-align: right;\"> -0.0888075</td><td style=\"text-align: right;\">  0.230126 </td><td style=\"text-align: right;\"> -0.118062 </td><td style=\"text-align: right;\">  0.271298</td><td style=\"text-align: right;\"> -0.012716  </td><td style=\"text-align: right;\"> -0.103982 </td><td style=\"text-align: right;\"> -0.0813083 </td><td style=\"text-align: right;\"> -0.128282</td><td style=\"text-align: right;\"> -0.108122  </td><td style=\"text-align: right;\"> -0.0775247 </td><td style=\"text-align: right;\"> -0.164759 </td><td style=\"text-align: right;\">  0.0308948 </td><td style=\"text-align: right;\"> -0.142819</td><td style=\"text-align: right;\"> -0.0529479</td><td style=\"text-align: right;\"> -0.109634 </td><td style=\"text-align: right;\">  0.00985533</td><td style=\"text-align: right;\">  0.110615   </td><td style=\"text-align: right;\">  0.00188479</td><td style=\"text-align: right;\"> -0.0144661</td><td style=\"text-align: right;\">  0.144588 </td><td style=\"text-align: right;\">  0.0140727 </td><td style=\"text-align: right;\">  0.183001  </td><td style=\"text-align: right;\">  0.102967  </td><td style=\"text-align: right;\">  0.0359119  </td><td style=\"text-align: right;\">  0.0288003</td><td style=\"text-align: right;\">  0.0220569 </td></tr>\n",
       "<tr><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan         </td><td style=\"text-align: right;\">nan         </td><td style=\"text-align: right;\">nan         </td><td style=\"text-align: right;\">nan         </td><td style=\"text-align: right;\">nan       </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan         </td><td style=\"text-align: right;\">nan         </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan          </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan         </td><td style=\"text-align: right;\">nan         </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan       </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan         </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan          </td><td style=\"text-align: right;\">nan         </td><td style=\"text-align: right;\">nan         </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan       </td><td style=\"text-align: right;\">nan         </td><td style=\"text-align: right;\">nan         </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan         </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan       </td><td style=\"text-align: right;\">nan          </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan       </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan         </td><td style=\"text-align: right;\">nan       </td><td style=\"text-align: right;\">nan       </td><td style=\"text-align: right;\">nan         </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan       </td><td style=\"text-align: right;\">nan       </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan          </td><td style=\"text-align: right;\">nan         </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan       </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan       </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan       </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan         </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan          </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan       </td><td style=\"text-align: right;\">nan         </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan         </td><td style=\"text-align: right;\">nan       </td><td style=\"text-align: right;\">nan         </td><td style=\"text-align: right;\">nan         </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan         </td><td style=\"text-align: right;\">nan       </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan         </td><td style=\"text-align: right;\">nan          </td><td style=\"text-align: right;\">nan         </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan         </td><td style=\"text-align: right;\">nan         </td><td style=\"text-align: right;\">nan         </td><td style=\"text-align: right;\">nan          </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan         </td></tr>\n",
       "<tr><td style=\"text-align: right;\"> -0.0397735</td><td style=\"text-align: right;\"> -0.183837 </td><td style=\"text-align: right;\">  0.00498708</td><td style=\"text-align: right;\"> -0.0225942 </td><td style=\"text-align: right;\">  0.105435  </td><td style=\"text-align: right;\">  0.0680924 </td><td style=\"text-align: right;\"> -0.153242</td><td style=\"text-align: right;\"> -0.0228845</td><td style=\"text-align: right;\"> -0.0288095 </td><td style=\"text-align: right;\"> -0.0585174 </td><td style=\"text-align: right;\">  0.0302902</td><td style=\"text-align: right;\"> -0.0662779</td><td style=\"text-align: right;\">  0.169481 </td><td style=\"text-align: right;\"> -0.0524676</td><td style=\"text-align: right;\">  0.0253732  </td><td style=\"text-align: right;\"> -0.0469438</td><td style=\"text-align: right;\"> -0.00186827</td><td style=\"text-align: right;\">  0.0691416 </td><td style=\"text-align: right;\">  0.0480555</td><td style=\"text-align: right;\">  0.0288044</td><td style=\"text-align: right;\">  0.21301 </td><td style=\"text-align: right;\"> -0.081737 </td><td style=\"text-align: right;\">  0.164588 </td><td style=\"text-align: right;\"> -0.100222  </td><td style=\"text-align: right;\">  0.0571208</td><td style=\"text-align: right;\"> -0.000977767</td><td style=\"text-align: right;\">  0.100034  </td><td style=\"text-align: right;\"> -0.102374  </td><td style=\"text-align: right;\">  0.0488952</td><td style=\"text-align: right;\">  0.282687</td><td style=\"text-align: right;\">  0.139644  </td><td style=\"text-align: right;\">  0.117263  </td><td style=\"text-align: right;\"> -0.137761 </td><td style=\"text-align: right;\">  0.116318 </td><td style=\"text-align: right;\"> -0.180992  </td><td style=\"text-align: right;\"> -0.0616165</td><td style=\"text-align: right;\">  0.108342 </td><td style=\"text-align: right;\">  0.13396  </td><td style=\"text-align: right;\">  0.0435324</td><td style=\"text-align: right;\">  0.0331406</td><td style=\"text-align: right;\"> -0.0483853</td><td style=\"text-align: right;\"> -0.0855508</td><td style=\"text-align: right;\"> -0.104852</td><td style=\"text-align: right;\"> -0.164283   </td><td style=\"text-align: right;\"> -0.0300262</td><td style=\"text-align: right;\"> -0.057    </td><td style=\"text-align: right;\"> -0.357296</td><td style=\"text-align: right;\"> -0.179854 </td><td style=\"text-align: right;\"> -0.00603578</td><td style=\"text-align: right;\"> -0.290705</td><td style=\"text-align: right;\">  0.133248</td><td style=\"text-align: right;\"> -0.03336   </td><td style=\"text-align: right;\">  0.0663261</td><td style=\"text-align: right;\">  0.122618 </td><td style=\"text-align: right;\"> -0.147496</td><td style=\"text-align: right;\">  0.247903</td><td style=\"text-align: right;\"> -0.0381681</td><td style=\"text-align: right;\"> -0.00225705 </td><td style=\"text-align: right;\"> -0.00829265</td><td style=\"text-align: right;\"> -0.123877 </td><td style=\"text-align: right;\"> -0.215466</td><td style=\"text-align: right;\"> -0.045041 </td><td style=\"text-align: right;\">  0.150038 </td><td style=\"text-align: right;\">  0.147179</td><td style=\"text-align: right;\"> -0.0142328</td><td style=\"text-align: right;\">  0.176262</td><td style=\"text-align: right;\"> -0.0559384</td><td style=\"text-align: right;\"> -0.0919878</td><td style=\"text-align: right;\"> -0.0760059</td><td style=\"text-align: right;\"> -0.0914566 </td><td style=\"text-align: right;\"> -0.0169465</td><td style=\"text-align: right;\">  0.0561966</td><td style=\"text-align: right;\"> -0.0407839  </td><td style=\"text-align: right;\"> -0.0691648</td><td style=\"text-align: right;\"> -0.137967 </td><td style=\"text-align: right;\">  0.0940351</td><td style=\"text-align: right;\">  0.0334881</td><td style=\"text-align: right;\">  0.142244</td><td style=\"text-align: right;\"> -0.00487407</td><td style=\"text-align: right;\">  0.0583538</td><td style=\"text-align: right;\">  0.0121206 </td><td style=\"text-align: right;\"> -0.13006 </td><td style=\"text-align: right;\"> -0.0109958 </td><td style=\"text-align: right;\"> -0.129883  </td><td style=\"text-align: right;\"> -0.0736615</td><td style=\"text-align: right;\">  0.0883417 </td><td style=\"text-align: right;\"> -0.186718</td><td style=\"text-align: right;\">  0.07145  </td><td style=\"text-align: right;\"> -0.0648334</td><td style=\"text-align: right;\"> -0.0971438 </td><td style=\"text-align: right;\">  0.157076   </td><td style=\"text-align: right;\">  0.144708  </td><td style=\"text-align: right;\"> -0.0985316</td><td style=\"text-align: right;\">  0.110975 </td><td style=\"text-align: right;\">  0.0510709 </td><td style=\"text-align: right;\"> -0.00653997</td><td style=\"text-align: right;\">  0.00867736</td><td style=\"text-align: right;\">  0.182193   </td><td style=\"text-align: right;\"> -0.0496458</td><td style=\"text-align: right;\">  0.0551999 </td></tr>\n",
       "<tr><td style=\"text-align: right;\"> -0.0459897</td><td style=\"text-align: right;\"> -0.0839729</td><td style=\"text-align: right;\">  0.0758659 </td><td style=\"text-align: right;\">  0.102955  </td><td style=\"text-align: right;\">  0.0498879 </td><td style=\"text-align: right;\">  0.0535371 </td><td style=\"text-align: right;\"> -0.162047</td><td style=\"text-align: right;\">  0.175324 </td><td style=\"text-align: right;\">  0.0578942 </td><td style=\"text-align: right;\"> -0.044408  </td><td style=\"text-align: right;\">  0.0748782</td><td style=\"text-align: right;\"> -0.0976649</td><td style=\"text-align: right;\"> -0.0544583</td><td style=\"text-align: right;\"> -0.0179788</td><td style=\"text-align: right;\"> -0.0238187  </td><td style=\"text-align: right;\"> -0.0638613</td><td style=\"text-align: right;\"> -0.0729243 </td><td style=\"text-align: right;\"> -0.00926253</td><td style=\"text-align: right;\"> -0.0180535</td><td style=\"text-align: right;\"> -0.167198 </td><td style=\"text-align: right;\">  0.168944</td><td style=\"text-align: right;\"> -0.0170811</td><td style=\"text-align: right;\">  0.097202 </td><td style=\"text-align: right;\">  0.00151923</td><td style=\"text-align: right;\">  0.0945489</td><td style=\"text-align: right;\">  0.0692719  </td><td style=\"text-align: right;\">  0.00169883</td><td style=\"text-align: right;\"> -0.00260909</td><td style=\"text-align: right;\">  0.0760248</td><td style=\"text-align: right;\">  0.180164</td><td style=\"text-align: right;\"> -0.023396  </td><td style=\"text-align: right;\">  0.0804647 </td><td style=\"text-align: right;\"> -0.0272891</td><td style=\"text-align: right;\">  0.150093 </td><td style=\"text-align: right;\"> -0.109454  </td><td style=\"text-align: right;\"> -0.0206549</td><td style=\"text-align: right;\">  0.0252643</td><td style=\"text-align: right;\">  0.0480099</td><td style=\"text-align: right;\"> -0.137857 </td><td style=\"text-align: right;\">  0.0149671</td><td style=\"text-align: right;\"> -0.0784054</td><td style=\"text-align: right;\"> -0.0985906</td><td style=\"text-align: right;\"> -0.203892</td><td style=\"text-align: right;\"> -0.078245   </td><td style=\"text-align: right;\"> -0.0265119</td><td style=\"text-align: right;\"> -0.0736557</td><td style=\"text-align: right;\"> -0.228983</td><td style=\"text-align: right;\"> -0.0898714</td><td style=\"text-align: right;\"> -0.0895335 </td><td style=\"text-align: right;\"> -0.141232</td><td style=\"text-align: right;\">  0.127505</td><td style=\"text-align: right;\"> -0.00641323</td><td style=\"text-align: right;\">  0.180669 </td><td style=\"text-align: right;\">  0.140742 </td><td style=\"text-align: right;\"> -0.11534 </td><td style=\"text-align: right;\">  0.186993</td><td style=\"text-align: right;\">  0.0231593</td><td style=\"text-align: right;\"> -0.0349532  </td><td style=\"text-align: right;\"> -0.163701  </td><td style=\"text-align: right;\"> -0.165272 </td><td style=\"text-align: right;\"> -0.187954</td><td style=\"text-align: right;\">  0.030701 </td><td style=\"text-align: right;\">  0.105644 </td><td style=\"text-align: right;\">  0.228515</td><td style=\"text-align: right;\"> -0.0596746</td><td style=\"text-align: right;\">  0.180261</td><td style=\"text-align: right;\"> -0.127441 </td><td style=\"text-align: right;\"> -0.0445718</td><td style=\"text-align: right;\"> -0.0447451</td><td style=\"text-align: right;\"> -0.062384  </td><td style=\"text-align: right;\"> -0.0512178</td><td style=\"text-align: right;\">  0.11672  </td><td style=\"text-align: right;\">  0.0709778  </td><td style=\"text-align: right;\"> -0.0712091</td><td style=\"text-align: right;\"> -0.146387 </td><td style=\"text-align: right;\">  0.178098 </td><td style=\"text-align: right;\"> -0.0929125</td><td style=\"text-align: right;\">  0.204458</td><td style=\"text-align: right;\">  0.0749398 </td><td style=\"text-align: right;\"> -0.0447788</td><td style=\"text-align: right;\">  0.0147287 </td><td style=\"text-align: right;\"> -0.118635</td><td style=\"text-align: right;\"> -0.019632  </td><td style=\"text-align: right;\"> -0.0886888 </td><td style=\"text-align: right;\"> -0.123245 </td><td style=\"text-align: right;\"> -0.00471   </td><td style=\"text-align: right;\"> -0.211728</td><td style=\"text-align: right;\"> -0.0701255</td><td style=\"text-align: right;\"> -0.115573 </td><td style=\"text-align: right;\"> -0.0752895 </td><td style=\"text-align: right;\">  0.0366755  </td><td style=\"text-align: right;\">  0.0583063 </td><td style=\"text-align: right;\"> -0.0758366</td><td style=\"text-align: right;\">  0.0786248</td><td style=\"text-align: right;\"> -0.0479902 </td><td style=\"text-align: right;\">  0.11444   </td><td style=\"text-align: right;\">  0.0757244 </td><td style=\"text-align: right;\">  0.0195613  </td><td style=\"text-align: right;\">  0.0169133</td><td style=\"text-align: right;\">  0.0651879 </td></tr>\n",
       "<tr><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan         </td><td style=\"text-align: right;\">nan         </td><td style=\"text-align: right;\">nan         </td><td style=\"text-align: right;\">nan         </td><td style=\"text-align: right;\">nan       </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan         </td><td style=\"text-align: right;\">nan         </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan          </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan         </td><td style=\"text-align: right;\">nan         </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan       </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan         </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan          </td><td style=\"text-align: right;\">nan         </td><td style=\"text-align: right;\">nan         </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan       </td><td style=\"text-align: right;\">nan         </td><td style=\"text-align: right;\">nan         </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan         </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan       </td><td style=\"text-align: right;\">nan          </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan       </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan         </td><td style=\"text-align: right;\">nan       </td><td style=\"text-align: right;\">nan       </td><td style=\"text-align: right;\">nan         </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan       </td><td style=\"text-align: right;\">nan       </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan          </td><td style=\"text-align: right;\">nan         </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan       </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan       </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan       </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan         </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan          </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan       </td><td style=\"text-align: right;\">nan         </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan         </td><td style=\"text-align: right;\">nan       </td><td style=\"text-align: right;\">nan         </td><td style=\"text-align: right;\">nan         </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan         </td><td style=\"text-align: right;\">nan       </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan         </td><td style=\"text-align: right;\">nan          </td><td style=\"text-align: right;\">nan         </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan         </td><td style=\"text-align: right;\">nan         </td><td style=\"text-align: right;\">nan         </td><td style=\"text-align: right;\">nan          </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan         </td></tr>\n",
       "<tr><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan         </td><td style=\"text-align: right;\">nan         </td><td style=\"text-align: right;\">nan         </td><td style=\"text-align: right;\">nan         </td><td style=\"text-align: right;\">nan       </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan         </td><td style=\"text-align: right;\">nan         </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan          </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan         </td><td style=\"text-align: right;\">nan         </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan       </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan         </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan          </td><td style=\"text-align: right;\">nan         </td><td style=\"text-align: right;\">nan         </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan       </td><td style=\"text-align: right;\">nan         </td><td style=\"text-align: right;\">nan         </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan         </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan       </td><td style=\"text-align: right;\">nan          </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan       </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan         </td><td style=\"text-align: right;\">nan       </td><td style=\"text-align: right;\">nan       </td><td style=\"text-align: right;\">nan         </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan       </td><td style=\"text-align: right;\">nan       </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan          </td><td style=\"text-align: right;\">nan         </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan       </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan       </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan       </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan         </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan          </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan       </td><td style=\"text-align: right;\">nan         </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan         </td><td style=\"text-align: right;\">nan       </td><td style=\"text-align: right;\">nan         </td><td style=\"text-align: right;\">nan         </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan         </td><td style=\"text-align: right;\">nan       </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan         </td><td style=\"text-align: right;\">nan          </td><td style=\"text-align: right;\">nan         </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan         </td><td style=\"text-align: right;\">nan         </td><td style=\"text-align: right;\">nan         </td><td style=\"text-align: right;\">nan          </td><td style=\"text-align: right;\">nan        </td><td style=\"text-align: right;\">nan         </td></tr>\n",
       "</tbody>\n",
       "</table>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/plain": []
     },
     "execution_count": 28,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train_desc_vecs"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "# Add aggregated word embeddings \n",
    "ext_train = train.cbind(train_desc_vecs)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "The word embeddings have added 100 columns to our data."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Number of columns in original data: 15\n",
      "Number of columns with word embeddings: 115\n"
     ]
    }
   ],
   "source": [
    "print(\"Number of columns in original data: \" + str(train.ncol))\n",
    "print(\"Number of columns with word embeddings: \" + str(ext_train.ncol))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We will also add the word embeddings to our test data.  We have trained our Word2Vec model on the train data only.  This is not necessary since we are training an unsupervised learning model on the `desc` column only so there should be limited data leakage.  By building a Word2Vec model on the train data and applying this model to the test data, however, we are most closely mimicking how this process will be done in production."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "test_desc = tokenize(test[\"desc\"].ascharacter())\n",
    "test_desc_vecs = w2v_model.transform(test_desc, aggregate_method=\"AVERAGE\")\n",
    "ext_test = test.cbind(test_desc_vecs)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We will now train our model with word embeddings as new features in replace of `desc`."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "gbm Model Build progress: |███████████████████████████████████████████████| 100%\n"
     ]
    }
   ],
   "source": [
    "# Train GBM model with word embeddings\n",
    "\n",
    "predictors = list(set(ext_train.col_names) - set([target, \"desc\"]))\n",
    "\n",
    "gbm_word_embeddings = H2OGradientBoostingEstimator(seed = 1234, model_id = \"gbm_word-embeddings.hex\", \n",
    "                                                   ntrees = 500, \n",
    "                                                   stopping_rounds = 10, score_tree_interval = 5, # early stopping\n",
    "                                                   stopping_metric = \"AUC\", \n",
    "                                                   nfolds = 5 # 5-fold cross validation\n",
    "                                                  )\n",
    "gbm_word_embeddings.train(x = predictors, y = target, training_frame = ext_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 60,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "AUC on CV - Baseline: 0.719\n",
      "AUC on CV - No Desc: 0.72077\n",
      "AUC on CV - Word Embeddings: 0.72084\n"
     ]
    }
   ],
   "source": [
    "print(\"AUC on CV - Baseline: \" + str(round(gbm_baseline.auc(xval = True), 4)))\n",
    "print(\"AUC on CV - No Desc: \" + str(round(gbm_no_desc.auc(xval = True), 5)))\n",
    "print(\"AUC on CV - Word Embeddings: \" + str(round(gbm_word_embeddings.auc(xval = True), 5)))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "By adding the word embeddings of `desc`, we have slightly better performance than removing `desc` all together.  The improvement is small, however, which could indicate that the description of the loan is not very informative in predicting whether or not a loan will be fully paid off.\n",
    "\n",
    "The variable importance plot shows us that the word embeddings do have some importance to the model."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAA3EAAAJTCAYAAACmZ5/FAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nAGBEhDgAAIAREeIAAABG5P8DhBYEqaVBG18AAAAASUVORK5CYII=\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x1201f5cf8>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "gbm_word_embeddings.varimp_plot(num_of_features=20)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Step 4. Perform Target Encoding on Categorical Features\n",
    "\n",
    "From our original baseline model, we can see that the variable `addr_state` has very high importance. It is surprising that `addr_state` has such high variable importance. The high variable importance could be because our model is memorizing the training data through this high cardinality categorical column. See: [Best Practices - Categorical Predictors](https://github.com/h2oai/h2o-tutorials/tree/master/best-practices/categorical-predictors) for more information on why this occurs."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAA3EAAAJTCAYAAACmZ5/FAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n7D64VEj6k6wAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x1203b5b70>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "gbm_baseline.varimp_plot()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "If we train our same GBM model without `addr_state` we actually see an improvement in model performance on the holdout data. This indicates that we are indeed overfitting on the high cardinality column.  "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "gbm Model Build progress: |███████████████████████████████████████████████| 100%\n"
     ]
    }
   ],
   "source": [
    "# Train GBM model without state\n",
    "\n",
    "gbm_no_state = H2OGradientBoostingEstimator(seed = 1234, model_id = \"gbm_no-state.hex\", \n",
    "                                            ntrees = 500, \n",
    "                                            stopping_rounds = 10, score_tree_interval = 5, # early stopping\n",
    "                                            stopping_metric = \"AUC\", \n",
    "                                            nfolds = 5 # 5-fold cross validation\n",
    "                                           )\n",
    "gbm_no_state.train(x = list(set(predictors) - set([\"addr_state\"])), y = target, training_frame = ext_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "AUC on CV - Word Embeddings: 0.721\n",
      "AUC on CV - No State: 0.726\n"
     ]
    }
   ],
   "source": [
    "print(\"AUC on CV - Word Embeddings: \" + str(round(gbm_word_embeddings.auc(xval = True), 3)))\n",
    "print(\"AUC on CV - No State: \" + str(round(gbm_no_state.auc(xval = True), 3)))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "To prevent this overfitting, we can perform target encoding on the `addr_state` column and use that in place of the original categorical column.  Target encoding is a powerful method of converting a categorical column to numeric.  Target Encoding is the practice of replacing a categorical value with the mean of the response.  For example, we can replace NY in our dataset with the probability of default for the loans originating in NY.  \n",
    "\n",
    "Note: Our `TargetEncoder` function includes many options to prevent overfitting.  We will use cross validation Target Encoding for this example.  For more information on Target Encoding, see our documentation: http://docs.h2o.ai/h2o/latest-stable/h2o-docs/data-munging/target-encoding.html"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# Add Fold Column for Target Encoding\n",
    "ext_train[\"cv_fold_te\"] = ext_train.kfold_column(n_folds = 5, seed = 54321)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<h2o.expr.H2OCache at 0x12048f048>"
      ]
     },
     "execution_count": 39,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from h2o.targetencoder import TargetEncoder\n",
    "\n",
    "# Fit target encoding on training data\n",
    "targetEncoder = TargetEncoder(x= [\"addr_state\", \"purpose\"], y = \"bad_loan\", fold_column = \"cv_fold_te\")\n",
    "targetEncoder.fit(ext_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "# Apply to train and test data\n",
    "ext_train = targetEncoder.transform(frame=ext_train, \n",
    "                                    holdout_type=\"kfold\", \n",
    "                                    seed=54321, \n",
    "                                    is_train_or_valid=True,\n",
    "                                    noise = 0\n",
    "                                   )\n",
    "\n",
    "ext_test = targetEncoder.transform(frame=ext_test, \n",
    "                                   holdout_type=\"none\", # no need to do holdout because \n",
    "                                                        #target encoder was calculated on train\n",
    "                                   seed=54321, \n",
    "                                   is_train_or_valid=False,\n",
    "                                   noise = 0\n",
    "                                  )"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Now for each Fold ID and `addr_state` we have a numeric value.  We will use the target encoding representation of `addr_state` instead of the original column in our next model.  We have done the same step for `purpose` which is another categorical feature with a high number of categories."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<table>\n",
       "<thead>\n",
       "<tr><th style=\"text-align: right;\">  cv_fold_te</th><th>addr_state  </th><th style=\"text-align: right;\">  addr_state_te</th><th>purpose  </th><th style=\"text-align: right;\">  purpose_te</th></tr>\n",
       "</thead>\n",
       "<tbody>\n",
       "<tr><td style=\"text-align: right;\">           0</td><td>AK          </td><td style=\"text-align: right;\">       0.276596</td><td>car      </td><td style=\"text-align: right;\">    0.145078</td></tr>\n",
       "<tr><td style=\"text-align: right;\">           0</td><td>AL          </td><td style=\"text-align: right;\">       0.195572</td><td>car      </td><td style=\"text-align: right;\">    0.145078</td></tr>\n",
       "<tr><td style=\"text-align: right;\">           0</td><td>AR          </td><td style=\"text-align: right;\">       0.144828</td><td>car      </td><td style=\"text-align: right;\">    0.145078</td></tr>\n",
       "<tr><td style=\"text-align: right;\">           0</td><td>AZ          </td><td style=\"text-align: right;\">       0.197441</td><td>car      </td><td style=\"text-align: right;\">    0.145078</td></tr>\n",
       "<tr><td style=\"text-align: right;\">           0</td><td>AZ          </td><td style=\"text-align: right;\">       0.197441</td><td>car      </td><td style=\"text-align: right;\">    0.145078</td></tr>\n",
       "<tr><td style=\"text-align: right;\">           0</td><td>AZ          </td><td style=\"text-align: right;\">       0.197441</td><td>car      </td><td style=\"text-align: right;\">    0.145078</td></tr>\n",
       "<tr><td style=\"text-align: right;\">           0</td><td>AZ          </td><td style=\"text-align: right;\">       0.197441</td><td>car      </td><td style=\"text-align: right;\">    0.145078</td></tr>\n",
       "<tr><td style=\"text-align: right;\">           0</td><td>AZ          </td><td style=\"text-align: right;\">       0.197441</td><td>car      </td><td style=\"text-align: right;\">    0.145078</td></tr>\n",
       "<tr><td style=\"text-align: right;\">           0</td><td>CA          </td><td style=\"text-align: right;\">       0.20982 </td><td>car      </td><td style=\"text-align: right;\">    0.145078</td></tr>\n",
       "<tr><td style=\"text-align: right;\">           0</td><td>CA          </td><td style=\"text-align: right;\">       0.20982 </td><td>car      </td><td style=\"text-align: right;\">    0.145078</td></tr>\n",
       "</tbody>\n",
       "</table>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/plain": []
     },
     "execution_count": 43,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "ext_train[[\"cv_fold_te\", \"addr_state\", \"addr_state_te\", \"purpose\", \"purpose_te\"]]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 44,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "gbm Model Build progress: |███████████████████████████████████████████████| 100%\n"
     ]
    }
   ],
   "source": [
    "# Train GBM model using target encoding instead of state and purpose\n",
    "\n",
    "predictors = list(set(ext_train.col_names) - set([\"cv_fold_te\", target, \"desc\", \"addr_state\", \"purpose\"]))\n",
    "\n",
    "gbm_te = H2OGradientBoostingEstimator(seed = 1234, model_id = \"gbm_target-encoding.hex\", \n",
    "                                      ntrees = 500, \n",
    "                                      stopping_rounds = 10, score_tree_interval = 5, # early stopping\n",
    "                                      stopping_metric = \"AUC\",\n",
    "                                      nfolds = 5 # 5-fold cross validation\n",
    "                                     )\n",
    "gbm_te.train(x = predictors, y = target, training_frame = ext_train)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "By adding the target encoding of `addr_state` and `purpose` and removing the original high cardinality features from the predictors, we have very similar performance to the model where we removed `addr_state` completely.  This could indicate that the state and purpose of the loan is not very informative in predicting whether or not a loan will be fully paid off."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 58,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "AUC on CV - Word Embeddings: 0.721\n",
      "AUC on CV - No State: 0.726\n",
      "AUC on CV - Target Encoding: 0.726\n"
     ]
    }
   ],
   "source": [
    "print(\"AUC on CV - Word Embeddings: \" + str(round(gbm_word_embeddings.auc(xval = True), 3)))\n",
    "print(\"AUC on CV - No State: \" + str(round(gbm_no_state.auc(xval = True), 3)))\n",
    "print(\"AUC on CV - Target Encoding: \" + str(round(gbm_te.auc(xval = True), 3)))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "The variable `addr_state_te` has much smaller variable importance than the original `addr_state`.  This indicates that the model is no longer overfitting on `addr_state` information."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 46,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAA3EAAAJTCAYAAACmZ5/FAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nAABJRU5ErkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x120ef1898>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "gbm_te.varimp_plot()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Step 5. Add Anomaly Score\n",
    "\n",
    "We can also use unsupervised learning models to add interesting information about the data.  For example: \n",
    "\n",
    "* what are the natural segments found in the data\n",
    "* which records are strange\n",
    "\n",
    "In this step, we will use the Isolation Forest algorithm to add an anomaly score to each record.  We will see if this information improves the model performance."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 47,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "isolationforest Model Build progress: |███████████████████████████████████| 100%\n"
     ]
    }
   ],
   "source": [
    "# Train Isolation Forest\n",
    "\n",
    "from h2o.estimators import H2OIsolationForestEstimator\n",
    "\n",
    "isolation_forest = H2OIsolationForestEstimator(seed = 1234,\n",
    "                                               model_id = \"isolation_forest.hex\"\n",
    "                                              )\n",
    "isolation_forest.train(training_frame = ext_train, x = predictors)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "When you predict with isolation forest, the output is `predict` and `mean_length`.  The `mean_length` value is the average number of splits it takes to segment the record in a randomized tree.  The more splits it takes, the less likely the record is an anomaly.\n",
    "\n",
    "The `predict` value is the `mean_length` normalized between 0 and 1.  Note: there may be some values less than 0 or greater than 1.  This because the current implementation calculates the min and max for normalization on a sample of the records."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 48,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "isolationforest prediction progress: |████████████████████████████████████| 100%\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<table>\n",
       "<thead>\n",
       "<tr><th style=\"text-align: right;\">    predict</th><th style=\"text-align: right;\">  mean_length</th></tr>\n",
       "</thead>\n",
       "<tbody>\n",
       "<tr><td style=\"text-align: right;\"> 0.0294118 </td><td style=\"text-align: right;\">         6.78</td></tr>\n",
       "<tr><td style=\"text-align: right;\"> 0.0441176 </td><td style=\"text-align: right;\">         6.74</td></tr>\n",
       "<tr><td style=\"text-align: right;\">-0.0367647 </td><td style=\"text-align: right;\">         6.96</td></tr>\n",
       "<tr><td style=\"text-align: right;\">-0.00735294</td><td style=\"text-align: right;\">         6.88</td></tr>\n",
       "<tr><td style=\"text-align: right;\"> 0.227941  </td><td style=\"text-align: right;\">         6.24</td></tr>\n",
       "<tr><td style=\"text-align: right;\">-0.0367647 </td><td style=\"text-align: right;\">         6.96</td></tr>\n",
       "<tr><td style=\"text-align: right;\"> 0.0735294 </td><td style=\"text-align: right;\">         6.66</td></tr>\n",
       "<tr><td style=\"text-align: right;\"> 0.102941  </td><td style=\"text-align: right;\">         6.58</td></tr>\n",
       "<tr><td style=\"text-align: right;\">-0.0147059 </td><td style=\"text-align: right;\">         6.9 </td></tr>\n",
       "<tr><td style=\"text-align: right;\">-0.0220588 </td><td style=\"text-align: right;\">         6.92</td></tr>\n",
       "</tbody>\n",
       "</table>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/plain": []
     },
     "execution_count": 48,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "isolation_forest.predict(ext_train)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We will add the normalized Isolation Forest prediction to our training and testing data."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 49,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "isolationforest prediction progress: |████████████████████████████████████| 100%\n"
     ]
    }
   ],
   "source": [
    "# Add Isolation Forest scores to train \n",
    "anomaly_scores_train = isolation_forest.predict(ext_train)[\"predict\"]\n",
    "anomaly_scores_train.col_names = [\"AnomalyScore\"]\n",
    "ext_train = ext_train.cbind(anomaly_scores_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 50,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "isolationforest prediction progress: |████████████████████████████████████| 100%\n"
     ]
    }
   ],
   "source": [
    "# Add Isolation Forest scores to test \n",
    "anomaly_scores_test = isolation_forest.predict(ext_test)[\"predict\"]\n",
    "anomaly_scores_test.col_names = [\"AnomalyScore\"]\n",
    "ext_test = ext_test.cbind(anomaly_scores_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 51,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "gbm Model Build progress: |███████████████████████████████████████████████| 100%\n"
     ]
    }
   ],
   "source": [
    "# Train GBM model with anomaly score as new feature\n",
    "\n",
    "predictors = list(set(ext_train.col_names) - set([\"cv_fold_te\", target, \"desc\", \"addr_state\", \"purpose\", \n",
    "                                                 \"addr_state_te\", \"purpose_te\"]))\n",
    "\n",
    "gbm_anomaly_score = H2OGradientBoostingEstimator(seed = 1234, model_id = \"gbm_anomaly-scores.hex\", \n",
    "                                                 ntrees = 500, \n",
    "                                                 stopping_rounds = 10, score_tree_interval = 5, # early stopping\n",
    "                                                 stopping_metric = \"AUC\",\n",
    "                                                 nfolds = 5 # 5-fold cross validation\n",
    "                                                )\n",
    "gbm_anomaly_score.train(x = predictors, y = target, training_frame = ext_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 52,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "AUC on CV - Baseline: 0.719\n",
      "AUC on CV - Target Encoding: 0.726\n",
      "AUC on CV - Anomaly Score: 0.722\n"
     ]
    }
   ],
   "source": [
    "print(\"AUC on CV - Baseline: \" + str(round(gbm_baseline.auc(xval = True), 3)))\n",
    "print(\"AUC on CV - Target Encoding: \" + str(round(gbm_te.auc(xval = True), 3)))\n",
    "print(\"AUC on CV - Anomaly Score: \" + str(round(gbm_anomaly_score.auc(xval = True), 3)))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Step 6. Evaluate Improvement to Model\n",
    "\n",
    "Now that we have completed our feature engineering, we will evaluate how the model performs on our test data."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 53,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "AUC on CV - Baseline: 0.719\n",
      "AUC on CV - with Anomaly Score: 0.722\n"
     ]
    }
   ],
   "source": [
    "print(\"AUC on CV - Baseline: \" + str(round(gbm_baseline.auc(xval = True), 4)))\n",
    "print(\"AUC on CV - with Anomaly Score: \" + str(round(gbm_anomaly_score.auc(xval = True), 4)))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 61,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "AUC on Test - Baseline: 0.724\n",
      "AUC on Test - Baseline: 0.7325\n",
      "AUC on Test - with Anomaly Score: 0.7324\n"
     ]
    }
   ],
   "source": [
    "print(\"AUC on Test - Baseline: \" + str(round(gbm_baseline.model_performance(ext_test).auc(), 4)))\n",
    "print(\"AUC on Test - with Anomaly Score: \" + str(round(gbm_anomaly_score.model_performance(ext_test).auc(), 4)))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Our final model does significantly better on the test data."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Step 7. Analyze Final Model\n",
    "\n",
    "Our last step is to analyze our final model.\n",
    "\n",
    "The `issue_d` has the greatest importance.  We can use the partial dependency plot to understand the relationship between `issue_d` and the probability of default."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 55,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAA3EAAAJTCAYAAACmZ5/FAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\niEOcJEmSJDXEIU6SJEmSGvIv78+KO+ixZUQAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x12182ce48>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "gbm_anomaly_score.varimp_plot()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 56,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "PartialDependencePlot progress: |█████████████████████████████████████████| 100%\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAf4AAALWCAYAAACnePHjAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\niIgkiBK/iIhIgvw/GR55gNzA3fAAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x1201c3748>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "%matplotlib inline\n",
    "pdp = gbm_anomaly_score.partial_plot(data=ext_train, cols = [\"issue_d\"])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "The partial dependency plot for `issue_d` shows that there is a significant decrease in probability of loan default over time.  \n",
    "\n",
    "This variable is the most important feature.  The GBM model may extrapolate into the future so removing this variable should be considered."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 62,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "H2O session _sid_b2a5 closed.\n"
     ]
    }
   ],
   "source": [
    "h2o.cluster().shutdown()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.6.0"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}