data-scientist-salary.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {
    "focus": false,
    "id": "69b9a648-bcc7-490d-9f9b-ea244d156bd6"
   },
   "source": [
    "# Web Scraping for Indeed.com & Predicting Salaries"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "focus": false,
    "id": "34681254-c802-462f-829d-8894d0772d08"
   },
   "source": [
    "Indeed.com is a leading job search website, allowing job-seekers to apply either on Indeed's site or on company websites.\n",
    "\n",
    "The goal of this project was to scrape data from Indeed and use the data to determine what are the indicators of determing data scientists' salary.  Salary information on data science jobs were collected in various cities.  Then, factors such as location, title, and keywords on job descriptions were analyzed.  In order to account for a natural variance in job salaries, the prolem was turned into a classification problem predicting the range of salary."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "focus": false,
    "id": "a948d79c-5527-4c0d-ab23-f5d43ce72056"
   },
   "source": [
    "### Webscraping\n",
    "BeautifulSoup was used to build a webscraper. URLs are adjusted to access data for different cities."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 50,
   "metadata": {
    "collapsed": false,
    "focus": false,
    "id": "2efefc73-064a-482d-b3b5-ddf5508cb4ec"
   },
   "outputs": [],
   "source": [
    "import requests\n",
    "import re\n",
    "import bs4\n",
    "from bs4 import BeautifulSoup\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "from sklearn.cross_validation import KFold, StratifiedKFold, cross_val_score\n",
    "from sklearn.ensemble import RandomForestClassifier, RandomForestRegressor\n",
    "import matplotlib.pyplot as plt\n",
    "%matplotlib inline\n",
    "from sklearn.metrics import confusion_matrix, classification_report\n",
    "from sklearn import svm\n",
    "from sklearn.neighbors import KNeighborsClassifier\n",
    "from sklearn.linear_model import LogisticRegression\n",
    "from sklearn.grid_search import GridSearchCV\n",
    "import seaborn as sns"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "collapsed": true,
    "focus": false,
    "id": "a1af53c9-9090-494f-b82e-cadb60a54909"
   },
   "outputs": [],
   "source": [
    "# functions to extract company name, job title, and salary\n",
    "def extract_text(el):\n",
    "    if el:\n",
    "        return el.text.strip()\n",
    "    else:\n",
    "        return ''\n",
    "def extract_company_from_result(result):\n",
    "    return extract_text(result.find('span',{'class':'company'}))\n",
    "def extract_job_from_result(result):\n",
    "    return result.find('a',{'data-tn-element':'jobTitle'}).text.strip()\n",
    "def extract_salary_from_result(result):\n",
    "    salary_table = result.find('td', {'class' : 'snip'})\n",
    "    if salary_table:\n",
    "        snip = salary_table.find('nobr')\n",
    "        if snip:\n",
    "            return snip.text.strip()\n",
    "    return None"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "collapsed": false,
    "focus": false,
    "id": "04b0f9af-540e-402f-8292-81748707c676"
   },
   "outputs": [],
   "source": [
    "# the first {} is for the location and the second [] is for the page number for the job search\n",
    "url_template = \"http://www.indeed.com/jobs?q=data+scientist+%2420%2C000&l={}&start={}\"\n",
    "max_results_per_city = 100 # set max results to 100 for faster scraping\n",
    "\n",
    "rows = []\n",
    "\n",
    "for city in set(['New+York', 'Chicago', 'San+Francisco', 'Austin', 'Seattle', \n",
    "    'Los+Angeles', 'Philadelphia', 'Atlanta', 'Dallas', 'Pittsburgh', \n",
    "    'Portland', 'Phoenix', 'Denver', 'Houston', 'Miami', 'Washington%2C+DC']):\n",
    "    for start in range(0, max_results_per_city, 10):\n",
    "        r = requests.get(url_template.format(city, start))\n",
    "        soup = BeautifulSoup(r.content)\n",
    "        results = soup.findAll('div', { \"class\" : \"result\" })\n",
    "        for result in results:\n",
    "            if result:\n",
    "                row={}\n",
    "                # use functions created above to obtain result\n",
    "                row['job_title']=extract_job_from_result(result)\n",
    "                row['company']=extract_company_from_result(result)\n",
    "                row['salary']=extract_salary_from_result(result)\n",
    "                row['city'] = city\n",
    "                # append all results\n",
    "                rows.append(row)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>city</th>\n",
       "      <th>company</th>\n",
       "      <th>job_title</th>\n",
       "      <th>salary</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>Houston</td>\n",
       "      <td>CyberCoders</td>\n",
       "      <td>Data Scientist - Analytics Consultant - Oil &amp; ...</td>\n",
       "      <td>None</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>Houston</td>\n",
       "      <td>Arundo Analytics</td>\n",
       "      <td>Machine Learning Expert</td>\n",
       "      <td>None</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>Houston</td>\n",
       "      <td>Philips Lighting</td>\n",
       "      <td>Data Scientist / Machine Learning Scientist</td>\n",
       "      <td>None</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>Houston</td>\n",
       "      <td>The Hanover Company</td>\n",
       "      <td>Quantitative Analyst</td>\n",
       "      <td>None</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>Houston</td>\n",
       "      <td>Houston Methodist</td>\n",
       "      <td>Scientist</td>\n",
       "      <td>None</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "      city              company  \\\n",
       "0  Houston          CyberCoders   \n",
       "1  Houston     Arundo Analytics   \n",
       "2  Houston     Philips Lighting   \n",
       "3  Houston  The Hanover Company   \n",
       "4  Houston    Houston Methodist   \n",
       "\n",
       "                                           job_title salary  \n",
       "0  Data Scientist - Analytics Consultant - Oil & ...   None  \n",
       "1                            Machine Learning Expert   None  \n",
       "2        Data Scientist / Machine Learning Scientist   None  \n",
       "3                               Quantitative Analyst   None  \n",
       "4                                          Scientist   None  "
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# converted results into a dataframe\n",
    "data = pd.DataFrame.from_records(rows)\n",
    "data.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(2400, 4)"
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "data.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "city            0\n",
       "company         0\n",
       "job_title       0\n",
       "salary       2252\n",
       "dtype: int64"
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# look for null values in salary\n",
    "data.isnull().sum()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "# remove duplicate rows\n",
    "data=data.drop_duplicates()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "# create a new dataframe without null values\n",
    "salary_data = data[data.salary.notnull()]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# remove non-annual salary \n",
    "salary_data = salary_data[~(salary_data.salary.astype('str').str.contains('hr'))]\n",
    "salary_data = salary_data[~(salary_data.salary.astype('str').str.contains('hour'))]\n",
    "salary_data = salary_data[~(salary_data.salary.astype('str').str.contains('week'))]\n",
    "salary_data = salary_data[~(salary_data.salary.astype('str').str.contains('wk'))]\n",
    "salary_data = salary_data[~(salary_data.salary.astype('str').str.contains('month'))]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>city</th>\n",
       "      <th>company</th>\n",
       "      <th>job_title</th>\n",
       "      <th>salary</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>Houston</td>\n",
       "      <td>Avispa Technology</td>\n",
       "      <td>Data Scientist 5796057</td>\n",
       "      <td>$80,000 - $120,000 a year</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>10</th>\n",
       "      <td>Houston</td>\n",
       "      <td>DISYS</td>\n",
       "      <td>Principal Data Scientist</td>\n",
       "      <td>$135,000 a year</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>50</th>\n",
       "      <td>Houston</td>\n",
       "      <td>Baylor College of Medicine</td>\n",
       "      <td>Senior Quality Assurance / Quality Control (Re...</td>\n",
       "      <td>$60,998 a year</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>52</th>\n",
       "      <td>Houston</td>\n",
       "      <td>Oscar Technology</td>\n",
       "      <td>Big Data Analytics Manager</td>\n",
       "      <td>$160,000 a year</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>65</th>\n",
       "      <td>Houston</td>\n",
       "      <td>Baylor College of Medicine</td>\n",
       "      <td>Bioinformatics Programmer II</td>\n",
       "      <td>$72,120 a year</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "       city                     company  \\\n",
       "9   Houston           Avispa Technology   \n",
       "10  Houston                       DISYS   \n",
       "50  Houston  Baylor College of Medicine   \n",
       "52  Houston            Oscar Technology   \n",
       "65  Houston  Baylor College of Medicine   \n",
       "\n",
       "                                            job_title  \\\n",
       "9                              Data Scientist 5796057   \n",
       "10                           Principal Data Scientist   \n",
       "50  Senior Quality Assurance / Quality Control (Re...   \n",
       "52                         Big Data Analytics Manager   \n",
       "65                       Bioinformatics Programmer II   \n",
       "\n",
       "                       salary  \n",
       "9   $80,000 - $120,000 a year  \n",
       "10            $135,000 a year  \n",
       "50             $60,998 a year  \n",
       "52            $160,000 a year  \n",
       "65             $72,120 a year  "
      ]
     },
     "execution_count": 18,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "salary_data.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true,
    "focus": false,
    "id": "a0f701e0-80bd-40ba-9101-4535860c0968"
   },
   "outputs": [],
   "source": [
    "# a function for using regex to extract salary amount\n",
    "def string_to_float(phrase):\n",
    "    return re.findall('(\\d{0,3}\\,\\d{0,3})',phrase)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# apply string_to_float to all salaries\n",
    "salary_data['salary']=salary_data['salary'].apply(string_to_float)\n",
    "# remove commas in salaries\n",
    "salary_data.salary=[[i.replace(',','') for i in y] for y in salary_data.salary]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>city</th>\n",
       "      <th>company</th>\n",
       "      <th>job_title</th>\n",
       "      <th>salary</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>Houston</td>\n",
       "      <td>Avispa Technology</td>\n",
       "      <td>Data Scientist 5796057</td>\n",
       "      <td>[80000.0, 120000.0]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>10</th>\n",
       "      <td>Houston</td>\n",
       "      <td>DISYS</td>\n",
       "      <td>Principal Data Scientist</td>\n",
       "      <td>[135000.0]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>50</th>\n",
       "      <td>Houston</td>\n",
       "      <td>Baylor College of Medicine</td>\n",
       "      <td>Senior Quality Assurance / Quality Control (Re...</td>\n",
       "      <td>[60998.0]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>52</th>\n",
       "      <td>Houston</td>\n",
       "      <td>Oscar Technology</td>\n",
       "      <td>Big Data Analytics Manager</td>\n",
       "      <td>[160000.0]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>65</th>\n",
       "      <td>Houston</td>\n",
       "      <td>Baylor College of Medicine</td>\n",
       "      <td>Bioinformatics Programmer II</td>\n",
       "      <td>[72120.0]</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "       city                     company  \\\n",
       "9   Houston           Avispa Technology   \n",
       "10  Houston                       DISYS   \n",
       "50  Houston  Baylor College of Medicine   \n",
       "52  Houston            Oscar Technology   \n",
       "65  Houston  Baylor College of Medicine   \n",
       "\n",
       "                                            job_title               salary  \n",
       "9                              Data Scientist 5796057  [80000.0, 120000.0]  \n",
       "10                           Principal Data Scientist           [135000.0]  \n",
       "50  Senior Quality Assurance / Quality Control (Re...            [60998.0]  \n",
       "52                         Big Data Analytics Manager           [160000.0]  \n",
       "65                       Bioinformatics Programmer II            [72120.0]  "
      ]
     },
     "execution_count": 22,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# convert the datatype of salary to float\n",
    "salary_data.salary=[[float(i) for i in y] for y in salary_data.salary]\n",
    "salary_data.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# if there are 2 salaries (range), average them\n",
    "salary_data.salary=[np.average(y) for y in salary_data.salary]\n",
    "salary_data.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {
    "collapsed": false,
    "focus": false,
    "id": "783fd153-28ac-47ab-bfca-27e7c1de95b4"
   },
   "outputs": [],
   "source": [
    "# save salary data to csv\n",
    "salary_data.to_csv('indeed-scraped-job-postings.csv', index=False, encoding='utf8')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "focus": false,
    "id": "04563b69-f7b6-466f-9d65-fc62c9ddee6a"
   },
   "source": [
    "### Modeling"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "collapsed": false,
    "focus": false,
    "id": "588f9845-6143-4bcc-bfd1-85d45b79303d"
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>city</th>\n",
       "      <th>company</th>\n",
       "      <th>job_title</th>\n",
       "      <th>salary</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>Houston</td>\n",
       "      <td>Avispa Technology</td>\n",
       "      <td>Data Scientist 5796057</td>\n",
       "      <td>100000.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>Houston</td>\n",
       "      <td>DISYS</td>\n",
       "      <td>Principal Data Scientist</td>\n",
       "      <td>135000.0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "      city            company                 job_title    salary\n",
       "0  Houston  Avispa Technology    Data Scientist 5796057  100000.0\n",
       "1  Houston              DISYS  Principal Data Scientist  135000.0"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "## load scraped salaries\n",
    "df=pd.read_csv('dataset/indeed-scraped-job-postings.csv')\n",
    "df.head(2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array(['Houston', 'Phoenix', 'Chicago', 'SF', 'NY', 'Dallas',\n",
       "       'Philadelphia', 'Denver', 'LA', 'Pittsburgh', 'Miami', 'Atlanta',\n",
       "       'Seattle', 'Austin', 'Portland', 'DC'], dtype=object)"
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# rename cities with long name for simplicity\n",
    "def simple_city(x):\n",
    "    if x=='Washington%2C+DC':\n",
    "        return 'DC'\n",
    "    elif x=='San+Francisco':\n",
    "        return 'SF'\n",
    "    elif x=='New+York':\n",
    "        return 'NY'\n",
    "    elif x=='Los+Angeles':\n",
    "        return 'LA'\n",
    "    else:\n",
    "        return x\n",
    "df.city=df.city.apply(simple_city)\n",
    "df.city.unique()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 54,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.axes._subplots.AxesSubplot at 0xd61b080>"
      ]
     },
     "execution_count": 54,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAA2oAAAGCCAYAAACRozDNAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\npKImhBBCCCGEEJ3M/wAtymSdRU/U1gAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0xd16a710>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "#grouping salary by city to find average salary of each city\n",
    "barcity=df.groupby(['city'])['salary'].mean()\n",
    "barcity.sort(ascending=False)\n",
    "#graph of average salary in each city\n",
    "sns.set_style(\"whitegrid\")\n",
    "plt.rcParams['figure.figsize']=(13,6)\n",
    "sns.barplot(x=barcity.index, y=barcity.values)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "collapsed": false,
    "focus": false,
    "id": "c20d2498-151c-44c3-a453-3a333c79a0ac"
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>city</th>\n",
       "      <th>company</th>\n",
       "      <th>job_title</th>\n",
       "      <th>salary</th>\n",
       "      <th>salary_bin</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>Houston</td>\n",
       "      <td>Avispa Technology</td>\n",
       "      <td>Data Scientist 5796057</td>\n",
       "      <td>100000.0</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>Houston</td>\n",
       "      <td>DISYS</td>\n",
       "      <td>Principal Data Scientist</td>\n",
       "      <td>135000.0</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "      city            company                 job_title    salary  salary_bin\n",
       "0  Houston  Avispa Technology    Data Scientist 5796057  100000.0           1\n",
       "1  Houston              DISYS  Principal Data Scientist  135000.0           1"
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# create a binary value based on whether salary is above or below mean\n",
    "def sal_bin(x):\n",
    "    if x > df.salary.mean():\n",
    "        return 1\n",
    "    else:\n",
    "        return 0\n",
    "df['salary_bin']=df['salary'].apply(sal_bin)\n",
    "df.head(2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "collapsed": false,
    "focus": false,
    "id": "87a17d3d-b7f4-4747-9f75-f9af1d18a174"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.43137254902\n"
     ]
    }
   ],
   "source": [
    "# baseline accuracy using the average value of all salary binary\n",
    "print df.salary_bin.mean()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "focus": false,
    "id": "4fb29de2-5b98-474c-a4ad-5170b72b9aea"
   },
   "source": [
    "### Building a Random Forest model to predict high/low salary"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>city</th>\n",
       "      <th>company</th>\n",
       "      <th>job_title</th>\n",
       "      <th>salary</th>\n",
       "      <th>salary_bin</th>\n",
       "      <th>city__Austin</th>\n",
       "      <th>city__Chicago</th>\n",
       "      <th>city__DC</th>\n",
       "      <th>city__Dallas</th>\n",
       "      <th>city__Denver</th>\n",
       "      <th>city__Houston</th>\n",
       "      <th>city__LA</th>\n",
       "      <th>city__Miami</th>\n",
       "      <th>city__NY</th>\n",
       "      <th>city__Philadelphia</th>\n",
       "      <th>city__Phoenix</th>\n",
       "      <th>city__Pittsburgh</th>\n",
       "      <th>city__Portland</th>\n",
       "      <th>city__SF</th>\n",
       "      <th>city__Seattle</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>Houston</td>\n",
       "      <td>Avispa Technology</td>\n",
       "      <td>Data Scientist 5796057</td>\n",
       "      <td>100000.0</td>\n",
       "      <td>1</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>Houston</td>\n",
       "      <td>DISYS</td>\n",
       "      <td>Principal Data Scientist</td>\n",
       "      <td>135000.0</td>\n",
       "      <td>1</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "      city            company                 job_title    salary  salary_bin  \\\n",
       "0  Houston  Avispa Technology    Data Scientist 5796057  100000.0           1   \n",
       "1  Houston              DISYS  Principal Data Scientist  135000.0           1   \n",
       "\n",
       "   city__Austin  city__Chicago  city__DC  city__Dallas  city__Denver  \\\n",
       "0           0.0            0.0       0.0           0.0           0.0   \n",
       "1           0.0            0.0       0.0           0.0           0.0   \n",
       "\n",
       "   city__Houston  city__LA  city__Miami  city__NY  city__Philadelphia  \\\n",
       "0            1.0       0.0          0.0       0.0                 0.0   \n",
       "1            1.0       0.0          0.0       0.0                 0.0   \n",
       "\n",
       "   city__Phoenix  city__Pittsburgh  city__Portland  city__SF  city__Seattle  \n",
       "0            0.0               0.0             0.0       0.0            0.0  \n",
       "1            0.0               0.0             0.0       0.0            0.0  "
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# make dummy varibles for cities because cities are categorical values\n",
    "df=pd.concat([df, pd.get_dummies(df['city'], prefix='city_', drop_first=True)], axis=1)\n",
    "df.head(2)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### First look at how location affects data science salary."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Accuracy: 0.56 (+/- 0.08)\n"
     ]
    }
   ],
   "source": [
    "# predict and print accuracy using random forest\n",
    "X = df.loc[:,'city__Austin':'city__Seattle'].copy()\n",
    "y = df['salary_bin']\n",
    "cv = KFold(len(y), n_folds=3, shuffle=True, random_state=0)\n",
    "rf = RandomForestClassifier()\n",
    "scores = cross_val_score(rf, X, y, cv=cv)\n",
    "print \"Accuracy: %0.2f (+/- %0.2f)\" % (scores.mean(), scores.std() * 2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "{'bootstrap': True, 'min_samples_leaf': 3, 'min_samples_split': 3, 'criterion': 'gini', 'max_features': 10, 'max_depth': 3}\n"
     ]
    }
   ],
   "source": [
    "# run grid search to optimize random forest\n",
    "search_parameters = {\"max_depth\": [3, None],\n",
    "              \"max_features\": [1, 3, 10],\n",
    "              \"min_samples_split\": [1, 3, 10],\n",
    "              \"min_samples_leaf\": [1, 3, 10],\n",
    "              \"bootstrap\": [True, False],\n",
    "              \"criterion\": [\"gini\", \"entropy\"]}\n",
    "X=df.loc[:,'city__Austin':'city__Seattle'].copy()\n",
    "y = df['salary_bin']\n",
    "grid = GridSearchCV(RandomForestClassifier(), search_parameters, cv=3)\n",
    "for fold in cv:\n",
    "    X_train = X.iloc[fold[0]]\n",
    "    X_test  = X.iloc[fold[1]]\n",
    "    y_train = y.iloc[fold[0]]\n",
    "    y_test  = y.iloc[fold[1]]\n",
    "grid.fit(X_train, y_train)\n",
    "print grid.best_params_"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>0 PREDICTED</th>\n",
       "      <th>1 PREDICTED</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0 TRUE</th>\n",
       "      <td>16</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1 TRUE</th>\n",
       "      <td>13</td>\n",
       "      <td>4</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "        0 PREDICTED  1 PREDICTED\n",
       "0 TRUE           16            1\n",
       "1 TRUE           13            4"
      ]
     },
     "execution_count": 19,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# make a confusion matrix\n",
    "y_pred = grid.predict(X_test)\n",
    "cm = confusion_matrix(y_test, y_pred)\n",
    "cm_df = pd.DataFrame(cm, columns=(['0 PREDICTED','1 PREDICTED']), index=(['0 TRUE','1 TRUE']))\n",
    "cm_df"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {
    "collapsed": false,
    "focus": false,
    "id": "ddbc6159-6854-4ca7-857f-bfecdaf6d9c2"
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>city</th>\n",
       "      <th>company</th>\n",
       "      <th>job_title</th>\n",
       "      <th>salary</th>\n",
       "      <th>salary_bin</th>\n",
       "      <th>city__Austin</th>\n",
       "      <th>city__Chicago</th>\n",
       "      <th>city__DC</th>\n",
       "      <th>city__Dallas</th>\n",
       "      <th>city__Denver</th>\n",
       "      <th>...</th>\n",
       "      <th>city__Portland</th>\n",
       "      <th>city__SF</th>\n",
       "      <th>city__Seattle</th>\n",
       "      <th>contains_senior</th>\n",
       "      <th>contains_junior</th>\n",
       "      <th>contains_manager</th>\n",
       "      <th>contains_bd</th>\n",
       "      <th>contains_scientist</th>\n",
       "      <th>contains_analyst</th>\n",
       "      <th>contains_engineer</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>Houston</td>\n",
       "      <td>Avispa Technology</td>\n",
       "      <td>data scientist 5796057</td>\n",
       "      <td>100000.0</td>\n",
       "      <td>1</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>True</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>Houston</td>\n",
       "      <td>DISYS</td>\n",
       "      <td>principal data scientist</td>\n",
       "      <td>135000.0</td>\n",
       "      <td>1</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>True</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>2 rows × 27 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "      city            company                 job_title    salary  salary_bin  \\\n",
       "0  Houston  Avispa Technology    data scientist 5796057  100000.0           1   \n",
       "1  Houston              DISYS  principal data scientist  135000.0           1   \n",
       "\n",
       "   city__Austin  city__Chicago  city__DC  city__Dallas  city__Denver  \\\n",
       "0           0.0            0.0       0.0           0.0           0.0   \n",
       "1           0.0            0.0       0.0           0.0           0.0   \n",
       "\n",
       "         ...          city__Portland  city__SF  city__Seattle  \\\n",
       "0        ...                     0.0       0.0            0.0   \n",
       "1        ...                     0.0       0.0            0.0   \n",
       "\n",
       "   contains_senior  contains_junior  contains_manager  contains_bd  \\\n",
       "0            False            False             False        False   \n",
       "1            False            False             False        False   \n",
       "\n",
       "   contains_scientist  contains_analyst  contains_engineer  \n",
       "0                True             False              False  \n",
       "1                True             False              False  \n",
       "\n",
       "[2 rows x 27 columns]"
      ]
     },
     "execution_count": 20,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# lowercase all job titles to make them easy to analyze\n",
    "# identify job titles keywords scuh as 'senior', 'junior', 'big data', scientist', 'engineer', etc.\n",
    "def lower(x):\n",
    "    return x.lower()\n",
    "df['job_title']=df['job_title'].apply(lower)\n",
    "df['contains_senior'] = df['job_title'].str.contains('senior','sr')\n",
    "df['contains_junior'] = df['job_title'].str.contains('junior','jr')\n",
    "df['contains_manager'] = df['job_title'].str.contains('manager')\n",
    "df['contains_bd'] = df['job_title'].str.contains('big data')\n",
    "df['contains_scientist'] = df['job_title'].str.contains('scientist')\n",
    "df['contains_analyst'] = df['job_title'].str.contains('analyst')\n",
    "df['contains_engineer'] = df['job_title'].str.contains('engineer')\n",
    "df.head(2)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Now consider keywords in job title and location to determine whther salary is below or above average."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Accuracy: 0.75 (+/- 0.08)\n"
     ]
    }
   ],
   "source": [
    "X = df.loc[:,'city__Austin':].copy()\n",
    "y = df['salary_bin']\n",
    "# create 2 folds and run a random forest classifier\n",
    "cv = KFold(len(y), n_folds=2,shuffle=True)\n",
    "rf = RandomForestClassifier()\n",
    "scores = cross_val_score(rf, X, y, cv=cv)\n",
    "print \"Accuracy: %0.2f (+/- %0.2f)\" % (scores.mean(), scores.std() * 2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "{'bootstrap': True, 'min_samples_leaf': 1, 'min_samples_split': 10, 'criterion': 'gini', 'max_features': 1, 'max_depth': 3}\n"
     ]
    }
   ],
   "source": [
    "# run gridsearch to optimize above random forest classifier\n",
    "search_parameters = {\"max_depth\": [3, None],\n",
    "              \"max_features\": [1, 3, 10],\n",
    "              \"min_samples_split\": [1, 3, 10],\n",
    "              \"min_samples_leaf\": [1, 3, 10],\n",
    "              \"bootstrap\": [True, False],\n",
    "              \"criterion\": [\"gini\", \"entropy\"]}\n",
    "X = df.loc[:,'city__Austin':].copy()\n",
    "y = df['salary_bin']\n",
    "for fold in cv:\n",
    "    X_train = X.iloc[fold[0]]\n",
    "    X_test  = X.iloc[fold[1]]\n",
    "    y_train = y.iloc[fold[0]]\n",
    "    y_test  = y.iloc[fold[1]]\n",
    "grid = GridSearchCV(RandomForestClassifier(), search_parameters, cv=2)\n",
    "grid.fit(X_train, y_train)\n",
    "print grid.best_params_"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "             precision    recall  f1-score   support\n",
      "\n",
      "          0       0.45      1.00      0.62        23\n",
      "          1       0.00      0.00      0.00        28\n",
      "\n",
      "avg / total       0.20      0.45      0.28        51\n",
      "\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "C:\\Users\\Ouisu\\Anaconda2\\lib\\site-packages\\sklearn\\metrics\\classification.py:1074: UndefinedMetricWarning: Precision and F-score are ill-defined and being set to 0.0 in labels with no predicted samples.\n",
      "  'precision', 'predicted', average, warn_for)\n"
     ]
    }
   ],
   "source": [
    "#classification report\n",
    "y_pred = grid.predict(X_test)\n",
    "print classification_report(y_test, y_pred)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>0 PREDICTED</th>\n",
       "      <th>1 PREDICTED</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0 TRUE</th>\n",
       "      <td>23</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1 TRUE</th>\n",
       "      <td>28</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "        0 PREDICTED  1 PREDICTED\n",
       "0 TRUE           23            0\n",
       "1 TRUE           28            0"
      ]
     },
     "execution_count": 26,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#make confusion matrix\n",
    "cm = confusion_matrix(y_test, y_pred)\n",
    "cm_df = pd.DataFrame(cm, columns=(['0 PREDICTED','1 PREDICTED']), index=(['0 TRUE','1 TRUE']))\n",
    "cm_df"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### SVM"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Accuracy: 0.72 (+/- 0.06)\n"
     ]
    }
   ],
   "source": [
    "#using svm to predict salary binary\n",
    "X=df.loc[:,'city__Austin':].copy()\n",
    "y = df['salary_bin']\n",
    "clf = svm.SVC(kernel='linear', C=1)\n",
    "scores = cross_val_score(rf, X, y, cv=cv)\n",
    "print \"Accuracy: %0.2f (+/- %0.2f)\" % (scores.mean(), scores.std() * 2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "{'kernel': 'rbf', 'C': 1}\n"
     ]
    }
   ],
   "source": [
    "#gridsearch on svm\n",
    "search_parameters = {'kernel':('linear', 'rbf'), 'C':[1, 10]}\n",
    "X=df.loc[:,'city__Austin':].copy()\n",
    "y = df['salary_bin']\n",
    "for fold in cv:\n",
    "    X_train = X.iloc[fold[0]]\n",
    "    X_test  = X.iloc[fold[1]]\n",
    "    y_train = y.iloc[fold[0]]\n",
    "    y_test  = y.iloc[fold[1]]\n",
    "grid = GridSearchCV(svm.SVC(), search_parameters, cv=2)\n",
    "grid.fit(X_train, y_train)\n",
    "print grid.best_params_"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "             precision    recall  f1-score   support\n",
      "\n",
      "          0       0.45      1.00      0.62        23\n",
      "          1       0.00      0.00      0.00        28\n",
      "\n",
      "avg / total       0.20      0.45      0.28        51\n",
      "\n"
     ]
    }
   ],
   "source": [
    "#classification report\n",
    "y_pred = grid.predict(X_test)\n",
    "print classification_report(y_test, y_pred)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>0 PREDICTED</th>\n",
       "      <th>1 PREDICTED</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0 TRUE</th>\n",
       "      <td>23</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1 TRUE</th>\n",
       "      <td>28</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "        0 PREDICTED  1 PREDICTED\n",
       "0 TRUE           23            0\n",
       "1 TRUE           28            0"
      ]
     },
     "execution_count": 30,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#make confusion matrix\n",
    "cm = confusion_matrix(y_test, y_pred)\n",
    "cm_df = pd.DataFrame(cm, columns=(['0 PREDICTED','1 PREDICTED']), index=(['0 TRUE','1 TRUE']))\n",
    "cm_df"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## KNN"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Accuracy: 0.69 (+/- 0.12)\n"
     ]
    }
   ],
   "source": [
    "X=df.loc[:,'city__Austin':].copy()\n",
    "y = df['salary_bin']\n",
    "knn = KNeighborsClassifier()\n",
    "scores = cross_val_score(knn, X, y, cv=cv)\n",
    "print \"Accuracy: %0.2f (+/- %0.2f)\" % (scores.mean(), scores.std() * 2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "{'n_neighbors': 2, 'weights': 'uniform'}\n"
     ]
    }
   ],
   "source": [
    "#gridsearch on knn\n",
    "search_parameters = {'weights': ['uniform', 'distance'], 'n_neighbors': [1,2,3,4,5,6,7,8,9, 10]}\n",
    "X=df.loc[:,'city__Austin':].copy()\n",
    "y = df['salary_bin']\n",
    "for fold in cv:\n",
    "    X_train = X.iloc[fold[0]]\n",
    "    X_test  = X.iloc[fold[1]]\n",
    "    y_train = y.iloc[fold[0]]\n",
    "    y_test  = y.iloc[fold[1]]\n",
    "grid = GridSearchCV(KNeighborsClassifier(), search_parameters, cv=2)\n",
    "grid.fit(X_train, y_train)\n",
    "print grid.best_params_"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "             precision    recall  f1-score   support\n",
      "\n",
      "          0       0.49      1.00      0.66        23\n",
      "          1       1.00      0.14      0.25        28\n",
      "\n",
      "avg / total       0.77      0.53      0.43        51\n",
      "\n"
     ]
    }
   ],
   "source": [
    "#knn classification report\n",
    "y_pred = grid.predict(X_test)\n",
    "print classification_report(y_test, y_pred)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>0 PREDICTED</th>\n",
       "      <th>1 PREDICTED</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0 TRUE</th>\n",
       "      <td>23</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1 TRUE</th>\n",
       "      <td>24</td>\n",
       "      <td>4</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "        0 PREDICTED  1 PREDICTED\n",
       "0 TRUE           23            0\n",
       "1 TRUE           24            4"
      ]
     },
     "execution_count": 35,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#make confusion matrix\n",
    "cm = confusion_matrix(y_test, y_pred)\n",
    "cm_df = pd.DataFrame(cm, columns=(['0 PREDICTED','1 PREDICTED']), index=(['0 TRUE','1 TRUE']))\n",
    "cm_df"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Logistic Regression"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Accuracy: 0.66 (+/- 0.14)\n"
     ]
    }
   ],
   "source": [
    "X=df.loc[:,'city__Austin':].copy()\n",
    "y = df['salary_bin']\n",
    "lm = LogisticRegression(penalty='l2', C=1)\n",
    "scores = cross_val_score(lm, X, y, cv=cv)\n",
    "print \"Accuracy: %0.2f (+/- %0.2f)\" % (scores.mean(), scores.std() * 2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "{'penalty': 'l1', 'C': 0.001}\n"
     ]
    }
   ],
   "source": [
    "#gridsearch on logistic regression\n",
    "search_parameters = {'C': [0.001, 0.01, 0.1, 1, 10, 100, 1000], 'penalty': ['l1','l2']}\n",
    "X=df.loc[:,'city__Austin':].copy()\n",
    "y = df['salary_bin']\n",
    "for fold in cv:\n",
    "    X_train = X.iloc[fold[0]]\n",
    "    X_test  = X.iloc[fold[1]]\n",
    "    y_train = y.iloc[fold[0]]\n",
    "    y_test  = y.iloc[fold[1]]\n",
    "grid = GridSearchCV(LogisticRegression(), search_parameters, cv=2)\n",
    "grid.fit(X_train, y_train)\n",
    "print grid.best_params_"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "             precision    recall  f1-score   support\n",
      "\n",
      "          0       0.45      1.00      0.62        23\n",
      "          1       0.00      0.00      0.00        28\n",
      "\n",
      "avg / total       0.20      0.45      0.28        51\n",
      "\n"
     ]
    }
   ],
   "source": [
    "y_pred = grid.predict(X_test)\n",
    "print classification_report(y_test, y_pred)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>0 PREDICTED</th>\n",
       "      <th>1 PREDICTED</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0 TRUE</th>\n",
       "      <td>23</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1 TRUE</th>\n",
       "      <td>28</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "        0 PREDICTED  1 PREDICTED\n",
       "0 TRUE           23            0\n",
       "1 TRUE           28            0"
      ]
     },
     "execution_count": 39,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#make confusion matrix\n",
    "cm = confusion_matrix(y_test, y_pred)\n",
    "cm_df = pd.DataFrame(cm, columns=(['0 PREDICTED','1 PREDICTED']), index=(['0 TRUE','1 TRUE']))\n",
    "cm_df"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Random Forest Regressor\n",
    "Instead of treating the problem as a classification problem, treat it as a regression problem using salary dollar amount."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {
    "collapsed": false,
    "focus": false,
    "id": "56cc8854-d722-411d-a6c7-e86310710f67"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Mean squared error: 0.10 (+/- 0.08)\n"
     ]
    }
   ],
   "source": [
    "X=df.loc[:,'city__Austin':].copy()\n",
    "y = df['salary']\n",
    "rfr = RandomForestRegressor(random_state=0, n_estimators=100)\n",
    "scores = cross_val_score(rfr, X, y, cv=5)\n",
    "print \"Mean squared error: %0.2f (+/- %0.2f)\" % (scores.mean(), scores.std() * 2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 45,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAdoAAAEWCAYAAADBzlZgAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n3UYAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0xca76588>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.style.use('fivethirtyeight')\n",
    "#predicted salary vs. actual salary\n",
    "def do_plot(model, m=None):\n",
    "    for fold in cv:\n",
    "        X_train = X.iloc[fold[0]]\n",
    "        X_test  = X.iloc[fold[1]]\n",
    "        y_train = y.iloc[fold[0]]\n",
    "        y_test  = y.iloc[fold[1]]\n",
    "        model.fit(X_train, y_train)\n",
    "        y_pred = model.predict(X_test)\n",
    "\n",
    "        plt.scatter(y_test, y_pred, s=100,alpha=.5)\n",
    "        plt.text(.1,.8, \"R2: \"+str(m), fontsize=20)\n",
    "        \n",
    "do_plot(rfr, scores.mean().round(2))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 2",
   "language": "python",
   "name": "python2"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 2
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython2",
   "version": "2.7.11"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 0
}