tang_tanya_evaluation.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# MIE1512H Project\n",
    "\n",
    "### Topic: Predicting Station-Level Demand in Bike-Sharing Systems\n",
    "\n",
    "Name: Tanya Tang<br>\n",
    "Version: V3<br>\n",
    "Date: March 27, 2020"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "___\n",
    "### Project Overview"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "The goal of the paper this project is based on is to improve the effectiveness of station-level rebalancing activites within bike-sharing systems by obtaining more accurate minimum inventory levels per station. \n",
    "\n",
    "The authors used historical bike trip data along with hourly weather data to train several machine learning models. The model features fell within two categories, time-related features from the bike data (i.e. year, month, etc.) and weather-related features (i.e. precipitation, temperature, etc.). To reduce overfitting, the data is first transformed onto a reduced space using four different reduction techniques:\n",
    "1. No reduction, baseline\n",
    "2. Group all stations into one cluster\n",
    "3. Kmeans - cluster stations into k clusters\n",
    "4. Singular value decomposition\n",
    "\n",
    "After reducing the problem, four different prediction methods were tested:\n",
    "1. Linear regression\n",
    "2. Multi-layer perceptron\n",
    "3. Gradient boosted tree\n",
    "4. Random forest\n",
    "\n",
    "Once a prediction was made on the reduced problem, the reduction needed to be inverted to obtain the full problem predictions. Each reduction technique has its own specific inversion process. \n",
    "\n",
    "The solution techniques were tested and scored based on the performance of its generated minimum station inventory levels. \n",
    "\n",
    "In this project, a similar methodology will be followed, with modifications and additions on the feature construction aspect. Focusing on the city of San Francisco, historical bike data and weather data will be used in conjunction with crime data. Features from all three data sets will be constructed, and several prediction techniques will be tested. As will be further explained in later sections, the reduction/inversion step was omitted from this project due to its more theoretical nature. Thus, we tested three of the four prediction methods from the paper, linear regression, gradient-boosted tree, and random forest. As per feedback from V1, before working on the full dataset, this notebook first applies the three prediction methods we tested to the toy dataset prepared from V1. Then, we will prepare the full dataset in the same manner as V1 and test the three prediction models. After the modeling is complete, we will create dynamic inventory models and test against a constant inventory model. This will conclude the project and we can see if the dynamic model is better. "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "___\n",
    "### Project Versions/Timelines"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### V1\n",
    "\n",
    "CRISP-DM Tasks:\n",
    "* Data Understanding (collecting, describing, exploring, verifying)\n",
    "* Data Preparation on Test Scale (selecting, cleaning, constructing, integrating, formatting)\n",
    "\n",
    "Actual Timeline:\n",
    "* Looking up potential datasets and collecting from online sources/deciding which datasets to use (0.5 hr)\n",
    "* Exploring entries/schema of each chosen dataset to understand discrepancies and connections between datasets (1 hr)\n",
    "* Loading/cleaning raw data within each dataset, taking a closer look at which entries are relevant (2 hr)\n",
    "* Calculating connections between different datasets and joining tables to eventually form a single aggregated table containing all relevant information for a single entry (4 hrs)\n",
    "* Verifying data preparation steps (i.e. tables joined correctly, entries were not lost throughout the preparation process, no null values exist) (1 hr)\n",
    "\n",
    "#### V2\n",
    "\n",
    "CRISP-DM Tasks:\n",
    "* Data Preparation on Full Scale (selecting, cleaning, constructing, integrating, formatting)\n",
    "* Modeling (selecting techniques, generating test design, building, assessing)\n",
    "\n",
    "Planned Timeline:\n",
    "* Recreate data preparation tasks from V1 on full dataset (1 hr)\n",
    "* Research reduction/prediction techniques and how to apply them using python packages (1 hr)\n",
    "* *Implement reduction techniques (4 hr) (did not complete)*\n",
    "* Implement regression techniques (4 hr)\n",
    "* *Implement inversion techniques (4 hr) (did not complete)*\n",
    "* Test implemented models (1 hr)\n",
    "\n",
    "Actual Timeline:\n",
    "* Research prediction techniques and how to apply them using python package (0.5 hr)\n",
    "* Apply ML models to dataset prepared in V1 (should have been included in V1) (3 hr)\n",
    "* Recreate data preparation tasks from V1 on full dataset (2 hr)\n",
    "* Implement regression techniques (3 hr)\n",
    "* Apply implemented models to test data to make predictions (4 hr)\n",
    "    * NOTE: Linear regression on all stations takes ~10 min to complete, gradient-boosted tree on all stations takes ~1 hr to complete, random forest on all stations takes ~30 min to complete\n",
    "\n",
    "#### V3\n",
    "\n",
    "CRISP-DM Tasks:\n",
    "* Modeling (assessing)\n",
    "* Evaluation (evaluting results, reviewing process, determining next steps)\n",
    "\n",
    "Planned Timeline:\n",
    "* *Transform predictions into inventory model for test data time range (4 hr) (did not complete)*\n",
    "* Create visualizations and compare performance between models (3 hr)\n",
    "* Review project activites to verify correctness (1 hr)\n",
    "* Summarize project and list some potential next steps (1 hr)\n",
    "\n",
    "Actual Timeline:\n",
    "* Create visualizations and compare performance between models (2 hr)\n",
    "* Review project activities to verify correctness/summarize project and list some potential next steps (1 hr)\n",
    "\n",
    "#### F\n",
    "\n",
    "Planned Timeline:\n",
    "* Write report (6 hr)\n",
    "\n",
    "Actual Timeline:\n",
    "* Write report (6 hr)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "___\n",
    "### Modeling"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Linear Regression\n",
    "* The linear regression model will be chosen from 6 different models via cross validations, the three regularization parameters are chosen from $(0.3, 0,5, 0.7)$ and the two elastic net parameters are chosen from $(0.5, 0.8)$\n",
    "* These parameters were selected by taking the default value given by spark, and slightly varying that value\n",
    "* The models will be evaluated by the built in RegressionEvaluator class in spark, giving an $R^2$ value\n",
    "\n",
    "#### Gradient-Boosted Tree\n",
    "* The gradient-boosted tree model will be chosen from 2 different models via cross validations, the two maximum depth parameters are chosen from $(3, 5)$\n",
    "* These parameters were selected by taking the value from the paper (5) and giving a lower parameter option as we have not applied any reduction methods so our model is more prone to overfitting\n",
    "* The models will be evaluated by the built in RegressionEvaluator class in spark, giving an $R^2$ value\n",
    "\n",
    "#### Random Forest\n",
    "* The random forest model will be chosen from 2 different models via cross validations, the two forest size (number of trees) parameters are chosen from $(50, 100)$\n",
    "* These parameters were selected by taking the value from the paper (100), and giving a lower parameter option as we have not applied any reduction methods so our model is more prone to overgitting\n",
    "* The models will be evaluated by the built in RegressionEvaluator class in spark, giving an $R^2$ value"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "___\n",
    "### Evaluation"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "To evaluate the prediction models we created, we need to generate and test inventory intervals based on those models. The paper fits a Poisson distribution to the station data then calculates the inventory intervals using several equations. This extra step is very theoretical and the equations are quite complex, involving integrals and probability distributions. Thus, I will not be completing this step for this project. Instead, I will focus on creating visualizations to compare results between the three prediction models. "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "First, let's load all of the appropriate data generated from V2 and check that each prediction model has results for the same stations. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import os"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Add Java locations\n",
    "os.environ[\"JAVA_HOME\"] = \"/usr/lib/jvm/java-1.8.0-openjdk-amd64\"\n",
    "os.environ[\"JRE_HOME\"] = \"/usr/lib/jvm/java-1.8.0-openjdk-amd64/jre\"\n",
    "\n",
    "# Initialize spark\n",
    "import findspark\n",
    "findspark.init(\"/opt/spark\")\n",
    "import pyspark\n",
    "spark = pyspark.sql.SparkSession.builder.appName(\"appName\").getOrCreate()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "def find_string(key, files):\n",
    "    return list(filter(lambda x: key in x, files))[0]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "True\n",
      "True\n",
      "True\n",
      "True\n",
      "True\n"
     ]
    }
   ],
   "source": [
    "# import station names and check that all models have same station results\n",
    "gbtr_depart_rootdir = '/home/tanya/Documents/Code/python/bike_share_rebalancing/resources/gbtr/departure'\n",
    "station_names = [r for _, r, _ in os.walk(gbtr_depart_rootdir)]\n",
    "station_names = station_names[0]\n",
    "gbtr_arrival_rootdir = '/home/tanya/Documents/Code/python/bike_share_rebalancing/resources/gbtr/arrival'\n",
    "print(sorted(station_names) == sorted([r for _, r, _ in os.walk(gbtr_arrival_rootdir)][0]))\n",
    "rfr_depart_rootdir = '/home/tanya/Documents/Code/python/bike_share_rebalancing/resources/rfr/departure'\n",
    "print(sorted(station_names) == sorted([r for _, r, _ in os.walk(rfr_depart_rootdir)][0]))\n",
    "rfr_arrival_rootdir = '/home/tanya/Documents/Code/python/bike_share_rebalancing/resources/rfr/arrival'\n",
    "print(sorted(station_names) == sorted([r for _, r, _ in os.walk(rfr_arrival_rootdir)][0]))\n",
    "lr_depart_rootdir = '/home/tanya/Documents/Code/python/bike_share_rebalancing/resources/lr/departure'\n",
    "print(sorted(station_names) == sorted([r for _, r, _ in os.walk(lr_depart_rootdir)][0]))\n",
    "lr_arrival_rootdir = '/home/tanya/Documents/Code/python/bike_share_rebalancing/resources/lr/arrival'\n",
    "print(sorted(station_names) == sorted([r for _, r, _ in os.walk(lr_arrival_rootdir)][0]))\n",
    "data = {i:{} for i in station_names}\n",
    "\n",
    "# import filepaths and add to data object\n",
    "gbtr_depart_parquet_files = list(filter(lambda x: x[-4:] == 'json', [os.path.join(r, n) for r, _, f in os.walk(gbtr_depart_rootdir) for n in f]))\n",
    "gbtr_arrival_parquet_files = list(filter(lambda x: x[-4:] == 'json', [os.path.join(r, n) for r, _, f in os.walk(gbtr_arrival_rootdir) for n in f]))\n",
    "rfr_depart_parquet_files = list(filter(lambda x: x[-4:] == 'json', [os.path.join(r, n) for r, _, f in os.walk(rfr_depart_rootdir) for n in f]))\n",
    "rfr_arrival_parquet_files = list(filter(lambda x: x[-4:] == 'json', [os.path.join(r, n) for r, _, f in os.walk(rfr_arrival_rootdir) for n in f]))\n",
    "lr_depart_parquet_files = list(filter(lambda x: x[-4:] == 'json', [os.path.join(r, n) for r, _, f in os.walk(lr_depart_rootdir) for n in f]))\n",
    "lr_arrival_parquet_files = list(filter(lambda x: x[-4:] == 'json', [os.path.join(r, n) for r, _, f in os.walk(lr_arrival_rootdir) for n in f]))\n",
    "for name in station_names:\n",
    "    data[name]['gbtr_depart'] = find_string(name, gbtr_depart_parquet_files)\n",
    "    data[name]['gbtr_arrival'] = find_string(name, gbtr_arrival_parquet_files)\n",
    "    data[name]['rfr_depart'] = find_string(name, rfr_depart_parquet_files)\n",
    "    data[name]['rfr_arrival'] = find_string(name, rfr_arrival_parquet_files)\n",
    "    data[name]['lr_depart'] = find_string(name, lr_depart_parquet_files)\n",
    "    data[name]['lr_arrival'] = find_string(name, lr_arrival_parquet_files)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Now we need to read in the json files to spark. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "....................................................................................................................................................................................................................................................................................................................................................."
     ]
    }
   ],
   "source": [
    "for name in data.keys():\n",
    "    data[name]['gbtr_depart'] = spark.read.json(data[name]['gbtr_depart'])\n",
    "    data[name]['gbtr_arrival'] = spark.read.json(data[name]['gbtr_arrival'])\n",
    "    data[name]['rfr_depart'] = spark.read.json(data[name]['rfr_depart'])\n",
    "    data[name]['rfr_arrival'] = spark.read.json(data[name]['rfr_arrival'])\n",
    "    data[name]['lr_depart'] = spark.read.json(data[name]['lr_depart'])\n",
    "    data[name]['lr_arrival'] = spark.read.json(data[name]['lr_arrival'])\n",
    "    print('.', end='')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "First, we can compare the the R^2 values calculated in V2 using the validation data. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 67,
   "metadata": {},
   "outputs": [],
   "source": [
    "validations = '/home/tanya/Documents/Code/python/bike_share_rebalancing/resources/validations'\n",
    "files = [f for _, _, f in os.walk(validations)][0]\n",
    "files = list(filter(lambda x: 'txt' in x, files))\n",
    "all_R2 = {i: 0 for i in list(map(lambda x: x.replace('_validation.txt', ''), files))}\n",
    "for file in files:\n",
    "    read = open(f'{validations}/{file}', 'r')\n",
    "    R2 = []\n",
    "    for line in read:\n",
    "        fields = line.split(' ')\n",
    "        R2.append(float(fields[-1][0:-2]))\n",
    "    all_R2[file.replace('_validation.txt', '')] = np.mean(R2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 80,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAagAAAEYCAYAAAAJeGK1AAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4yLjEsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy+j8jraAAASIElEQVR4nO3de7AedX3H8fcHIhdvyCVSJ1yCGqei0qgH+EOljAiG0QFqYQSxA60UdaCdtjotWguIwxRbpzJ26EDkImoRqVpJNSOilJZWxBwwoMFBAnJJxssRECuCCH77x7PRp6eHXM/z7O8k79fMM9n97W+f3/ckO/nM7v7ObqoKSZJas13fBUiSNBMDSpLUJANKktQkA0qS1CQDSpLUpHl9FzBOe+yxRy1cuLDvMiRJQ26++eYfV9X86e3bVEAtXLiQycnJvsuQJA1Jcu9M7V7ikyQ1yYCSJDXJgJIkNcmAkiQ1yYCSJDXJgJIkNcmAkiQ1yYCSJDXJgJIkNWmbepLEnHL2LmMc6+HxjSVJG8kzKElSkwwoSVKTDChJUpO8B7UJFp7xxbGNdc9OYxtKkprkGZQkqUkGlCSpSQaUJKlJBpQkqUkGlCSpSQaUJKlJBpQkqUkGlCSpSQaUJKlJBpQkqUkGlCSpSQaUJKlJPixWmiVjfZjweW8Y21hSXzyDkiQ1yYCSJDXJgJIkNcl7UNJcdPYuYxzr4fGNJQ3xDEqS1CQDSpLUJANKktSkXgMqyZIkdyRZneSMGbYfkuSWJE8kOXbatieTrOw+y8ZXtSRpHHqbJJFke+AC4HBgDbAiybKqun2o233AycC7Z/iKR6tq8cgLlST1os9ZfAcBq6vqboAkVwJHA78OqKq6p9v2qz4KlCT1p89LfAuA+4fW13RtG2unJJNJvp7kmKfqlOTUrt/k1NTU5tYqSRqzuTxJYt+qmgDeApyf5AUzdaqqpVU1UVUT8+fPH2+FkqTN1mdArQX2Hlrfq2vbKFW1tvvzbuB64OWzWZwkqV99BtQKYFGS/ZLsABwPbNRsvCS7JtmxW94DeBVD964kSXNfbwFVVU8ApwPXAN8BrqqqVUnOSXIUQJIDk6wBjgMuSrKq2/3FwGSSW4F/B86bNvtPkjTH9fosvqpaDiyf1nbm0PIKBpf+pu/3NeBlIy9QktSbuTxJQpK0FTOgJElNMqAkSU0yoCRJTTKgJElNMqAkSU0yoCRJTTKgJElNMqAkSU0yoCRJTTKgJElNMqAkSU0yoCRJTTKgJElNMqAkSU0yoCRJTTKgJElNMqAkSU0yoCRJTTKgJElNMqAkSU0yoCRJTTKgJElNMqAkSU0yoCRJTTKgJElNMqAkSU0yoCRJTTKgJElNMqAkSU0yoCRJTTKgJElN6jWgkixJckeS1UnOmGH7IUluSfJEkmOnbTspyZ3d56TxVS1JGofeAirJ9sAFwJHA/sAJSfaf1u0+4GTgimn77gacBRwMHASclWTXUdcsSRqfPs+gDgJWV9XdVfU4cCVw9HCHqrqnqm4DfjVt39cD11bVg1X1EHAtsGQcRUuSxqPPgFoA3D+0vqZrm9V9k5yaZDLJ5NTU1GYVKkkav61+kkRVLa2qiaqamD9/ft/lSJI2Up8BtRbYe2h9r65t1PtKkuaAPgNqBbAoyX5JdgCOB5Zt5L7XAEck2bWbHHFE1yZJ2kr0FlBV9QRwOoNg+Q5wVVWtSnJOkqMAkhyYZA1wHHBRklXdvg8CH2AQciuAc7o2SdJWYl6fg1fVcmD5tLYzh5ZXMLh8N9O+lwKXjrRASVJvtvpJEpKkucmAkiQ1yYCSJDXJgJIkNcmAkiQ1yYCSJDXJgJIkNcmAkiQ1yYCSJDXJgJIkNcmAkiQ1yYCSJDXJgJIkNcmAkiQ1yYCSJDXJgJIkNcmAkiQ1yYCSJDXJgJIkNcmAkiQ1yYCSJDXJgJIkNcmAkiQ1yYCSJDXJgJIkNWm9AZVk+yRvT/KBJK+atu19oy1NkrQt29AZ1EXA7wIPAB9J8g9D2940sqokSdu8DQXUQVX1lqo6HzgYeGaSzyXZEcjoy5Mkbas2FFA7rFuoqieq6lRgJXAd8MxRFiZJ2rZtKKAmkywZbqiqc4DLgIWjKkqSpPUGVFW9taq+NEP7xVX1tNGVJUna1m3UNPMk24+6EEmShm0woJI8C7h6FIMnWZLkjiSrk5wxw/Ydk3y6235TkoVd+8IkjyZZ2X0uHEV9kqT+zFvfxiTPAz4PnDvbA3dnZRcAhwNrgBVJllXV7UPd3gY8VFUvTHI88EHgzd22u6pq8WzXJUlqw4bOoG4AzquqZSMY+yBgdVXdXVWPA1cCR0/rczRwebf8GeCwJE5vl6RtwIYC6iFgwYjGXgDcP7S+Zoaxft2nqp4AHgZ277btl+SbSf4jyWueapAkpyaZTDI5NTU1e9VLkkZqQwF1KHBkktPGUMum+D6wT1W9HPgL4Iokz56pY1UtraqJqpqYP3/+WIuUJG2+DU0zfwQ4Cnj5CMZeC+w9tL5X1zZjnyTzgF2AB6rqF1X1QFfjzcBdwItGUKMkqScbnMVXVU9W1SkjGHsFsCjJfkl2AI4Hpt/rWgac1C0fC1xXVZVk/rqp70meDywC7h5BjZKknmzW6zaSbJfkxC0ZuLundDpwDfAd4KqqWpXknCRHdd0uAXZPsprBpbx1U9EPAW5LspLB5Il3VNWDW1KPJKktG5pm/mzgNAaTFZYB1zIIlXcBtwL/vCWDV9VyYPm0tjOHlh8Djpthv88Cn92SsSVJbVtvQAGfYDCT70bgFOC9DJ5ifkxVrRxxbZKkbdiGAur5VfUygCQX85vZc4+NvDJJ0jZtQ/egfrluoaqeBNYYTpKkcdjQGdTvJPlptxxg5249QFXVjL97JEnSllpvQFWVTzGXJPVis6aZS5I0agaUJKlJBpQkqUkGlCSpSQaUJKlJBpQkqUkGlCSpSQaUJKlJBpQkqUkGlCSpSQaUJKlJBpQkqUkGlCSpSQaUJKlJBpQkqUkGlCSpSQaUJKlJBpQkqUkGlCSpSQaUJKlJBpQkqUkGlCSpSQaUJKlJBpQkqUkGlCSpSQaUJKlJBpQkqUm9BlSSJUnuSLI6yRkzbN8xyae77TclWTi07T1d+x1JXj/OuiVJo9dbQCXZHrgAOBLYHzghyf7Tur0NeKiqXgh8GPhgt+/+wPHAS4AlwD913ydJ2kr0eQZ1ELC6qu6uqseBK4Gjp/U5Gri8W/4McFiSdO1XVtUvqup7wOru+yRJW4l5PY69ALh/aH0NcPBT9amqJ5I8DOzetX992r4LZhokyanAqQD77LPPFhX8+eMX8JOf/GSLvmNjXc/VYxlnMNj14xtrK/axJc8Y21geH3PP+V/57tjG+rN5nxvbWM855jwWL148ku/uM6DGoqqWAksBJiYmaku+a1T/CJK2fid/6ZGxjXXovBvHNhYj/H+xz0t8a4G9h9b36tpm7JNkHrAL8MBG7itJmsP6DKgVwKIk+yXZgcGkh2XT+iwDTuqWjwWuq6rq2o/vZvntBywCvjGmuiVJY9DbJb7untLpwDXA9sClVbUqyTnAZFUtAy4BPpFkNfAggxCj63cVcDvwBHBaVT3Zyw8iSRqJXu9BVdVyYPm0tjOHlh8DjnuKfc8Fzh1pgZKk3vgkCUlSkwwoSVKTDChJUpMMKElSk7b6X9SVpG3O2Q/3XcGs8AxKktQkA0qS1CQDSpLUJANKktQkA0qS1CQDSpLUJANKktQkA0qS1CQDSpLUJANKktQkA0qS1CQDSpLUJANKktQkA0qS1CQDSpLUJANKktQkA0qS1CQDSpLUJANKktQkA0qS1CQDSpLUJANKktQkA0qS1CQDSpLUJANKktQkA0qS1CQDSpLUpF4CKsluSa5Ncmf3565P0e+krs+dSU4aar8+yR1JVnaf546veknSOPR1BnUG8NWqWgR8tVv/P5LsBpwFHAwcBJw1LchOrKrF3edH4yhakjQ+fQXU0cDl3fLlwDEz9Hk9cG1VPVhVDwHXAkvGVJ8kqWd9BdSeVfX9bvkHwJ4z9FkA3D+0vqZrW+ey7vLe3yTJUw2U5NQkk0kmp6amtrhwSdJ4zBvVFyf5CvBbM2z66+GVqqoktYlff2JVrU3yLOCzwB8AH5+pY1UtBZYCTExMbOo4kqSejCygqup1T7UtyQ+TPK+qvp/kecBM95DWAocOre8FXN9999ruz/9JcgWDe1QzBpQkaW7q6xLfMmDdrLyTgKtn6HMNcESSXbvJEUcA1ySZl2QPgCRPA94IfHsMNUuSxqivgDoPODzJncDrunWSTCS5GKCqHgQ+AKzoPud0bTsyCKrbgJUMzrQ+Ov4fQZI0SiO7xLc+VfUAcNgM7ZPAKUPrlwKXTuvzCPDKUdcoSeqXT5KQJDXJgJIkNcmAkiQ1yYCSJDXJgJIkNcmAkiQ1yYCSJDXJgJIkNcmAkiQ1yYCSJDXJgJIkNcmAkiQ1yYCSJDXJgJIkNcmAkiQ1yYCSJDWplxcWStK25p7z3tB3CXOOZ1CSpCYZUJKkJhlQkqQmGVCSpCYZUJKkJhlQkqQmGVCSpCYZUJKkJhlQkqQmGVCSpCalqvquYWySTAH39l3HRtoD+HHfRahZHh9an7l2fOxbVfOnN25TATWXJJmsqom+61CbPD60PlvL8eElPklSkwwoSVKTDKh2Le27ADXN40Prs1UcH96DkiQ1yTMoSVKTDChJUpMMqB4k2TPJFUnuTnJzkhuT/F6SQ5M8nGRlktuSfCXJc5P8Yde2MsnjSb7VLZ/X98+i0UvyZPfv/e0k/5bkOV37wiSPDh0bK5Ps0He9Gq8kP5uh7ewka7tj4vYkJ/RR25byHtSYJQnwNeDyqrqwa9sXOAr4FvDuqnpj1/63wONVddbQ/vcAE1U1l34JT1sgyc+q6pnd8uXAd6vq3CQLgS9U1Uv7rE/9Gj4+htrOBn5WVR9Ksgi4Gdi9qn7ZR42byzOo8Xstg9C5cF1DVd1bVf843KkLsmcBD425PrXtRmBB30Vo7qiqO4GfA7v2Xcummtd3AduglwC3rGf7a5KsBHYHHgHeO5aq1Lwk2wOHAZcMNb+gO14A/ruqTht/ZWpZklcAd1bVj/quZVN5BtWzJBckuTXJiq7phqpaXFV7A5cBf9djeWrDzl0I/QDYE7h2aNtd3fGy2HDSNH+eZBVwE3Bu38VsDgNq/FYBr1i30v2nchjw/x6UCCwDDhlTXWrXo1W1GNgXCGAQaWN8uKpeAvw+cEmSnfouaFMZUON3HbBTkncOtT39Kfq+Grhr9CVpLqiqnwN/CrwriZfntVGqahkwCZzUdy2byoN8zKqqkhwDfDjJXwJTDO41/VXXZd09qAAPA6f0U6laVFXfTHIbcAJwQ9/1qAlPT7JmaP0fZuhzDnBFko9W1a/GVNcWc5q5JKlJXuKTJDXJgJIkNcmAkiQ1yYCSJDXJgJIkNcmAkmZJkkryyaH1eUmmknxhE7/nniR7bGqfJDd1T6++rxt33RPOF27K+FIr/D0oafY8Arw0yc5V9ShwOLB2XINX1cEASU5m8MT708c1tjQKnkFJs2s58IZu+QTgU+s2JNktyee7d319PckBXfvuSb6cZFWSixn8kva6fd6a5BvdmdBF3QNjN0qS7ZLcmWT+0PrqJPOTfCzJhUkmk3w3ybpXvGyf5O+TrOjqfPuW/5VIm8eAkmbXlcDx3XPPDmDwoM513g98s6oOYPCU+o937WcB/9U9N+1fgX0AkrwYeDPwqu5ZfE8CJ25sId0TAz45tM/rgFuraqpbXwgcxCBQL+xqfhvwcFUdCBwI/HGS/Tb+x5dmj5f4pFlUVbd193xOYHA2NezVDB7cSVVd1505PZvBA4Hf1LV/Mcm6d4AdBrwSWDF4PRg7A5v6yoRLgauB84E/YvCE/HWu6kLsziR3A78NHAEckOTYrs8uwCLge5s4rrTFDChp9i0DPgQcyuC9XpsrDN68/J7N/YKquj/JD5O8lsHZ0vAZ2PTnnFU35p9U1TWbO6Y0W7zEJ82+S4H3V9W3prXfQBcQSQ4FflxVPwX+E3hL134kv3nz6VeBY5M8t9u2W5J9N6Oeixlc6vuXqnpyqP247r7UC4DnA3cA1wDvTPK0bswXJXnGZowpbTHPoKRZVlVrgI/MsOls4NLuaeQ/5zevP3g/8Knu5XJfA+7rvuf2JO8DvpxkO+CXDN4Fde8mlrSMwaW9y6a13wd8A3g28I6qeqybpLEQuCWD64pTwDGbOJ40K3yaubSVSzLB4OV1rxlq+xjwhar6TG+FSRvgGZS0FUtyBvBONmH2n9QKz6AkSU1ykoQkqUkGlCSpSQaUJKlJBpQkqUkGlCSpSf8Ldrt/Y/lPMc4AAAAASUVORK5CYII=\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "\n",
    "fig, ax = plt.subplots()\n",
    "x = [1, 2, 3]\n",
    "width = 0.2\n",
    "y1 = [all_R2['departure_gradient_boosted_forest'],\n",
    "      all_R2['departure_random_forest'],\n",
    "      all_R2['departure_linear_regression']]\n",
    "y2 = [all_R2['arrival_gradient_boosted_forest'],\n",
    "      all_R2['arrival_random_forest'],\n",
    "      all_R2['arrival_linear_regression']]\n",
    "ax.bar([i - width / 2 for i in x], y1, width, label='departures')\n",
    "ax.bar([i + width / 2 for i in x], y2, width, label='arrivals')\n",
    "ax.plot([0.8, 3.2], [0, 0], color='grey', linewidth=0.5)\n",
    "ax.set_xticks([1, 2, 3])\n",
    "ax.set_xticklabels(['GBT', 'RF', 'LR'])\n",
    "ax.set_xlabel('Model Type')\n",
    "ax.set_ylabel('R^2')\n",
    "fig.tight_layout()\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We can see that none of the models are doing particularly well. This may be due to an unbalanced distribution of data samples across stations. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 104,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "fig, ax = plt.subplots()\n",
    "x = np.arange(0, len(data))\n",
    "width = 1\n",
    "for i, name in enumerate(data.keys()):\n",
    "    ax.bar(x[i], data[name]['gbtr_depart'].count(), width, color='grey')\n",
    "ax.set_ylabel('Number of Data Samples')\n",
    "plt.tick_params(axis='x', which='both', bottom=False, labelbottom=False)\n",
    "fig.tight_layout()\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 105,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAX4AAAEWCAYAAABhffzLAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4yLjEsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy+j8jraAAAgAElEQVR4nO3deZgcVfn28e+dhUCAhC0gEGJYFFzYYgB9UdkVRAGRRRAERHH9gaKvBuUV1FdBFKOgqAFkUVAEZBMF2XGDEEggwQCBECCsAdlXgef3xzltOp3JpKamqyczdX+uq66prup+6vQyT58+deocRQRmZlYfg/q6AGZm1llO/GZmNePEb2ZWM078ZmY148RvZlYzTvxmZjUzpK8LUMQqq6wSY8eO7etimJn1KzfffPPjETGqdXu/SPxjx45lypQpfV0MM7N+RdJ9XW13U4+ZWc048ZuZ1YwTv5lZzTjxm5nVjBO/mVnNOPGbmdWME7+ZWc048ZuZ1Uy/uIDLBraxEy4tdL85x+5ccUnM6sE1fjOzmnHiNzOrGSd+M7OaceI3M6sZn9y1AcknjM0WzTV+M7OaceI3M6sZJ34zs5px4jczqxmf3DUryCeMbaBwjd/MrGac+M3MasaJ38ysZtzGP4C5TdrMuuIav5lZzTjxm5nVjBO/mVnNOPGbmdWME7+ZWc1UnvglDZY0VdIf8+21Jd0o6W5J50haquoymJnZfJ2o8R8GzGy6/X1gYkSsBzwJHNyBMpiZWVZp4pc0GtgZOCXfFrAtcF6+yxnAblWWwczMFlR1jf/HwFeB1/PtlYGnIuLVfHsusGZXD5R0iKQpkqbMmzev4mKamdVHZYlf0geBxyLi5jKPj4hJETE+IsaPGjWqzaUzM6uvKods2BLYRdIHgKWBEcBPgBUkDcm1/tHAgxWWwczMWlRW44+IIyJidESMBT4KXB0RHwOuAfbIdzsAuKiqMpiZ2cL6oh//14DDJd1NavM/tQ/KYGZWWx0ZnTMirgWuzeuzgc07cdz+xCNpmlmn+MpdM7OaceI3M6sZJ34zs5px4jczqxknfjOzmnHiNzOrGSd+M7OaceI3M6sZJ34zs5px4jczqxknfjOzmnHiNzOrGSd+M7OaceI3M6sZJ34zs5px4jczqxknfjOzmlnsDFyStgSmRcTzkvYDxgE/iYj7Ki+dLXE8U5hZ/1ekxv9z4AVJGwNfBu4Bzqy0VGZmVpkiif/ViAhgV+CnEfEzYPlqi2VmZlUpMtn6s5KOAPYD3itpEDC02mKZmVlVitT49wZeBg6OiEeA0cAPKi2VmZlVZrE1/pzsf9R0+35q3sbvE5xm1p8ttsYvaXdJsyQ9LekZSc9KeqYThTMzs/Yr0sZ/HPChiJhZdWHMzKx6Rdr4H3XSNzMbOIrU+KdIOge4kHSSF4CI+ENlpTIzs8oUSfwjgBeA9zVtC8CJ38ysHyrSq+egThTEzMw6o0ivntGSLpD0WF7OlzS6E4UzM7P2K3Jy9zTgYmCNvFySt5mZWT9UJPGPiojTIuLVvJwOjKq4XGZmVpEiif8JSftJGpyX/YAnqi6YmZlVo0ji/wSwF/AI8DCwB+ATvmZm/VSRXj33Abt0oCxmZtYBi0z8kr4aEcdJOpHUb38BEXFopSUzM7NKdFfjbwzTMKUTBTEzs85YZOKPiEvy6gsRcW7zPkl7Li6wpKWB64Fh+TjnRcRRktYGfgesDNwM7B8Rr5Qsv5mZ9VCRk7tHFNzW6mVg24jYGNgE2FHSO4HvAxMjYj3gSeDgooU1M7Pe666NfyfgA8Cakk5o2jUCeHVxgfM8vc/lm0PzEsC2wL55+xnA0aQJ3c3MrAO6q/E/RGrff4nUJNNYLgbeXyR47vc/DXgMuAK4B3gqIhpfHHOBNRfx2EMkTZE0Zd68eUUOZ2ZmBXTXxn8rcKuksyPiP2WCR8RrwCaSVgAuADbowWMnAZMAxo8fv1CvIjMzK6fIsMxjJR0DvBVYurExItYpepCIeErSNcC7gBUkDcm1/tHAgz0ss5mZ9ULRQdp+TmrX34Y00fpvFvcgSaNyTR9JywA7kLqIXkO6+hfgAOCinhfbzMzKKpL4l4mIqwBFxH0RcTSwc4HHrQ5cI+k24Cbgioj4I/A14HBJd5O6dJ5aruhmZlZGkaaelyUNAmZJ+gKpaWa5xT0oIm4DNu1i+2xg854W1MzM2qNIjf8wYDhwKPAOYD/g41UWyszMqlMk8Y+NiOciYm5EHBQRHwHGVF0wMzOrRpVX7pqZ2RKosit3zcxsydTdyd3Glbu7kK7YbXgW+FKVhTIzs+oUvnJX0lDg7cCDEfFkx0poZmZttcg2fkm/kPS2nPRHAreSLt6aKmmfjpXQzMzaqruTu++JiNvz+kHAXRGxIalL51crL5mZmVWiu8TfPDnKDsCFABHxSKUlMjOzSnWX+J+S9EFJmwJbApcBSBoCLNOJwpmZWft116vn08AJwBuALzbV9LcDLq26YGZmVo3uevXcBezYxfbLgcurLJSZmVWnyJW7ZmY2gDjxm5nVTHf9+A/Lf7fsXHHMzKxq3dX4D8p/T+xEQczMrDO669UzU9IsYI08i1aDgIiIjaotmpmZVaG7Xj37SHoDqQfPLp0rkpmZVanbqRdz3/2NJS0FvDlvvjMi/lN5yczMrBKLnXNX0lakwdnmkJp51pJ0QERcX3HZzMysAkUmW/8R8L6IuBNA0puB35IGazMzs36mSD/+oY2kD/+9ondodUUyM7MqFanxT5F0CvCbfPtjpJm5zMysHyqS+D8LfB44NN/+K3BSZSUyM7NKLTbxR8TLpHb+H1VfHDMzq5rH6jEzqxknfjOzmlls4pe0YScKYmZmnVGkxn+SpMmSPidpZOUlMjOzSi028UfEe0hdONcCbpZ0tqQdKi+ZmZlVolAbf0TMAo4EvgZsBZwg6Q5Ju1dZODMza78ibfwbSZoIzAS2BT4UEW/J6xMrLp+ZmbVZkQu4TgROAb4eES82NkbEQ5KOrKxkZmZWiSKJf2fgxYh4DUDSIGDpiHghIn5daenMzKztirTxXwks03R7eN5mZmb9UJHEv3REPNe4kdeHV1ckMzOrUpHE/7ykcY0bkt4BvNjN/c3MbAlWpI3/i8C5kh4izcD1BmDvxT1I0lqkmbtWAwKYFBE/kbQScA4wljSr114R8WSp0puZWY8VGZ3zJkkbAOvnTUXn3H0V+HJE3CJpedLFX1cABwJXRcSxkiYAE0jXB5iZWQcUqfEDbEaqoQ8BxkkiIs7s7gER8TDwcF5/VtJMYE1gV2DrfLczgGtx4jcz65gik63/GlgXmAa8ljcHqRmnEEljgU2BG4HV8pcCwCOkpiAzM+uQIjX+8cBbIyLKHEDScsD5wBcj4hlJ/90XESGpy7iSDgEOARgzZkyZQ5st8cZOuLTQ/eYcu3PFJbE6KdKrZwbphG6PSRpKSvpnRcQf8uZHJa2e968OPNbVYyNiUkSMj4jxo0aNKnN4MzPrQpEa/yrAvyRNBl5ubIyIXbp7kFLV/lRgZkQ0T9t4MXAAcGz+e1FPC21mZuUVSfxHl4y9JbA/MF3StLzt66SE/3tJBwP3AXuVjG9mZiUU6c55naQ3Am+KiCslDQcGF3jc30j9/ruyXc+KWZ7bUM3MFlRkWOZPAecBv8yb1gQurLJQZmZWnSIndz9ParZ5Bv47KcuqVRbKzMyqUyTxvxwRrzRuSBpC6sdvZmb9UJHEf52krwPL5Ll2zwUuqbZYZmZWlSKJfwIwD5gOfBr4E2n+XTMz64eK9Op5HTg5L2Zm1s8VGavnXrpo04+IdSopkZmZVaroWD0NSwN7AitVUxwzM6vaYtv4I+KJpuXBiPgxaQJ2MzPrh4o09YxrujmI9Aug6Dj+Zma2hCmSwI9vWn+VPF1iJaUxM7PKFenVs00nCmJmZp1RpKnn8O72twy5bGZmS7iivXo2I42jD/AhYDIwq6pCmZlZdYok/tHAuIh4FkDS0cClEbFflQUzM7NqFBmyYTXglabbr+AJ0s3M+q0iNf4zgcmSLsi3dwPOqK5IZmZWpSK9er4r6c/Ae/KmgyJiarXFMjOzqhRp6gEYDjwTET8B5kpau8IymZlZhYpMvXgU8DXgiLxpKPCbKgtlZmbVKVLj/zCwC/A8QEQ8BCxfZaHMzKw6RRL/KxER5KGZJS1bbZHMzKxKRRL/7yX9ElhB0qeAK/GkLGZm/Va3vXokCTgH2AB4Blgf+GZEXNGBspmZWQW6TfwREZL+FBEbAk72ZmYDQJGmnlskbVZ5SczMrCOKXLm7BbCfpDmknj0i/RjYqMqCmZlZNRaZ+CWNiYj7gfd3sDxmZlax7mr8F5JG5bxP0vkR8ZFOFcrMzKrTXRu/mtbXqbogZmbWGd0l/ljEupmZ9WPdNfVsLOkZUs1/mbwO80/ujqi8dGZm1naLTPwRMbiTBTEzs84oOiyzmZkNEE78ZmY148RvZlYzTvxmZjVTWeKX9CtJj0ma0bRtJUlXSJqV/65Y1fHNzKxrVdb4Twd2bNk2AbgqIt4EXJVvm5lZB1WW+CPieuDfLZt3Bc7I62cAu1V1fDMz61qn2/hXi4iH8/ojwGodPr6ZWe312cnd5nl8uyLpEElTJE2ZN29eB0tmZjawdTrxPyppdYD897FF3TEiJkXE+IgYP2rUqI4V0MxsoOt04r8YOCCvHwBc1OHjm5nVXpXdOX8L/BNYX9JcSQcDxwI7SJoFbJ9vm5lZBxWZerGUiNhnEbu2q+qYZgZjJ1xa6H5zjt254pLYkspX7pqZ1YwTv5lZzTjxm5nVjBO/mVnNOPGbmdWME7+ZWc048ZuZ1YwTv5lZzTjxm5nVjBO/mVnNOPGbmdWME7+ZWc048ZuZ1YwTv5lZzTjxm5nVjBO/mVnNOPGbmdWME7+ZWc1UNvWimQ0cns5xYHGN38ysZpz4zcxqxonfzKxm3MZvZn3C5w36jmv8ZmY148RvZlYzTvxmZjXjxG9mVjNO/GZmNePEb2ZWM078ZmY148RvZlYzTvxmZjXjxG9mVjNO/GZmNePEb2ZWM078ZmY149E5zWzA6A8jfhYtI1RXzj6p8UvaUdKdku6WNKEvymBmVlcdT/ySBgM/A3YC3grsI+mtnS6HmVld9UWNf3Pg7oiYHRGvAL8Ddu2DcpiZ1VJfJP41gQeabs/N28zMrAMUEZ09oLQHsGNEfDLf3h/YIiK+0HK/Q4BD8s31gTvbWIxVgMfbGM8xHdMxHXNJjLl+RCzfurEvevU8CKzVdHt03raAiJgETKqiAJKmRMR4x3RMx3TMgR6zq+190dRzE/AmSWtLWgr4KHBxH5TDzKyWOl7jj4hXJX0BuBwYDPwqIm7vdDnMzOqqTy7giog/AX/qi2NnVTQhOaZjOqZj9ouYHT+5a2Zmfctj9ZiZ1YwTv5lZzdQy8Us6s6/LYGbWVwb86JySWruKCthG0goAEbFLm483KSIOWfw9+4aklSLi331djkWRtCLwWkQ808aY6wEbAzMj4l/tiptj9/r1lCTSUCaNK9gfBCZHyRNwkjaKiNt6U6aCx/lcRJxUQdzlIuK5ko8dAzwTEU9JGguMB+6IiBltLGK/V4ca/2jgGeBHwPF5ebZpvcckrbSIZWXgA2ULKmlM4wtJ0lhJe0h6ey/iHdm0/lZJdwE3S5ojaYuSMTeUdIOkByRNyom6sW9yyZhrSDpT0tOkKxdnSLpf0tGShpaId42kVfL6/qQeZDsB50j6nzJlzLG2lDRT0u2StpB0BXBTfi3eVTLm+4BZwNGkz84HgG8Bs/K+MqZKmiXpO+0aAFHS4S3Ll4FvN2634xhNSn0555F+rwNukPRJ4DLmv+9tK2O+Bml3SRv0Ms4ISet2sX2j3sQtJCIG9EL6cvsScAWwSd42u5cxXwNmA/c2LY3br5SMOSE//g7gk/nvqcDtwOElY97StH4psFNe3xz4R8mYfwN2BFYAvpLLt27eN7VkzKuBrfP67sBEYFng/wOTSsSb0bR+E7ByXh8O3NaL930ysCHwLtIX1Lvz9nHA30vGnAmM7WL72qRfKGViTgXeDnwXuBu4NX++FjpOD2I+C5wDfBM4Ki9PNtZLxDt8EcuXgX+XLOPtwDLAyrm8o/L2ZZs/EyXiXti0vmv+Pz2NNIzMgSVj7gU8BEzL5d6sad8tZcta+PhVH2BJWUg1/3OBnwL39zLWLGDMIvY9UDJm2z+0LYl/asu+skn61pbb2+TX451lP7BdxLy5af2OEvGmAmvm9WuApfP6YOD2XrzvU5vWZ7bsK/vcZwFDuti+FGkU21697/n25qRfvHMp/4U/Jv//fB8YnreVrkABLwHfafoSaV6eKhnztqb3+TFgUNO+3iT+5vf9H8DaeX2V1s9uD2JOA1Zven/uAD7ceryqlgHfxt8QEXOBPSXtTGr66Y0fAysC93ex77iSMV+LiBclvQK8CDwBEBHPpybgUtbJ5zgEjJY0PCJeyPt63ITSIGlkRDydy3eNpI8A5wMrlQw5T9J+pCS9OzAnH0eUa478EvAXSeeTvlCvlnQ58G5STa2s5rIc0bJvqZIxf0VqLvod80etXYs0lMmpJWMu8IGJiMnA5Nw8894yASPiftL/z67AFZImlixbwy2kmvTNrTtyM02pmJLOJlWWrgLOkHQZsC0lm4+y5nMtQyLiXoCIeFzS6yVjDo6Ih3OcyZK2Af4oaa2W41XCF3CVIGkYsDfwUERcKWlf4P+QfrZPioj/lIh5Oil5LAu8ALxKaqPcFlg+IvYqEXOrlk03R8RzklYD9oiIn5WIuS+ppndDy/YxwP+LiE+ViDkG+CFpYp5pwP+NiIfzOZOtI+L8EjFHAvsCbyZ1YpgLXBQRd/Q0VlPMXYArm748G9vXBT4SEaW+9HM7/C4seHL34ih5IlrSvhFxdpnHFoy/LOmcxBYRUeqLRNL6wBMRsdBolJJWi4hHS8QcAuxJSpznkWrS+5IqaD+LiOdLlvU14HnSF+ow4I3587kUMCUietwmL+kfwP4RcU/TthHABaQmxGFlylr4+E78PSfpLFIyGQ48BSwH/AHYjvSaHlAiZiUfWquWpFUj4rG+Lod1Xu6I8ZaI+GeJx24MvBARs1q2DwX2ioiz2lTMrlXdljQQF+a3JQ4BHiX9bINUIyh98rCCcr4B+DlpqsuVSbW06cDvye2LbT5ej0/E5setQ2ry+A7pS/RkYAapTXlsiXhfAFbJ6+sC15NORN4IbNiL57dSy7IyqVlqRWClkjFHAMcAvwb2adl3UsmYywHfJjVzPQ3MA26g5InIHPMW4Ejyifw2fFY2alofmmNfDHyPfA6hl6/lvu14LVtirEY6kT8OWK0N8YY0rS9H6npa6nPU06UO3TmrMCj/zFueVOsfmbcPo2TbuaQdm9ZXkHSqpNsknZ2bZso4ndS2+QCp/fxFUnfBvwK/KFnOKrqynk7qffM8KUHdQeqGdxnpC6GnPhvzmxBOACZGxIrA1yj5vLPHgZublimk5plb8noZp5EqDOeT5p8+PzclQjphXsZZpF5m7yd1DT0B2J90/cr3SsZckdST6xpJkyV9SdIaJWNBes8bjgXWI3WvXoby71Hza/nRNr2WSNpU0g3AtaRzeMcB1+VuzZuWjHkg8KikuyTtBNxGOnF+q6R9ypa1sE58uwy0hXTycDZwH3Ao6UTSyaTa9FElYzb3wDmF1JXxjflYF5aM2dwb4f6WfdNKxqyiK2t35exxDwfgzqb1m1r29aY755dJX0YbNm27t5efpWktt78B/J30a6JdvaRuyn8HUaKXVH5s8+fzPcBJwCOkCsUhvXzPpwFD83rpX81VvJZN5duii+3vbH2texBzOqlX0NqkziaNLtGr9eYzWnSpTa+edoqIiZLOyesPKQ0BsT1wcqQeFL01PiI2yesTJfX4nEHW/IuudZiKwSVjzga2i9TLYwGSHuji/kW8LunNpBrlcEnjI2KK0hW3Zcp5Xj5Z/m3gAklfJJ0025aue2IVEhHH5/d9Yn6uR9H7HhjDJA2KiNfzMb4r6UFS89RyJWM+L+ndEfG3fEL63zn26yrfRey/j4uIvwJ/VboYbgdSR4eeDik8UtLuOe6wyB0iIiIklX1Nq3gtAZaNiBtbN0bEDflEdxmvRfpV+rik5yKf5I2IR3vRi68wJ/6SIuKhpvWnSCdke2PVfHWhgBGSFLkKQPkrrC9Svvw9Ipqv4l2P8nMYV9GV9avAJcDrwG7AEfnqxZFAj3sJRcQ3JB0E/JbUxj+MNH/zhcDHSpaxEbvRLXgX0kWBw3sTj/S8twWubDrG6ZIeAU4sGfMzwCmS3kRq5/8EgKRRpPM9ZSz0eYmI10i/gC4rEe864IOkz/sNjZ48kt5A+Xlnq3gtAf4s6VJS5am5y+3HKffcAe6XdAypufgOSceTOohsDzzci7IW4l49SwhJR7VsOiki5uV/hOMi4uMl425Aaoe+MZrGP5G0Y0SU+tBK2pxUObspd0XckdSEUHpyHaUhJF7PMd9GauP/V29itsT/dUTs345YTTHfA2xFGlfnLyVjbEF67Z6WtAzp+oBNSedmvhf5eolexBxOOrcxrpcxDwUuiIiyv+pa4w0jXavwYLSpS3SOuw7pWpC1SM2SdwFnRy/Hfsrt8LuycJfbUp/P3HXz86RfjD8l/Q8dSKpQfSdyH/+qOPH3A5IOiogeX3iUf4p/gfTPtAlwWERclPfdEhHjSsQ8ipSUh5BqvFuQ2nl3AC6PiO+2IebmpBNppWJq4YH5INUEr4byA/NJmhwRm+f1TwGfI/2KeB9wSUQcWyLm7cDGkaYknUS6huM8UtfgjSNi9yUk5tOkk+/3kH5JnRsR83oapyled12iiYgDS8Q8lPQr4npSR4OpOfaHgc9FxLVlyzvgVH0SwUvvF0oOMUE6gbRcXh9L6nlyWL5ddsiG6aR29+Gkk1Ij8vZlKH9Srq0xSb1sfgNsTaqRb036+bwVsFUv3ofmE5I3seCwGtNLxpzZtN461ELZE/BVxJxKanJ8H+mK4nmkZo4DSBcY9jRe27tENz5HeX04cG1eH1P2854fP5LU82gm6XzJE3n9WGCFXsa8o10xe7K4O+cSInfd7GqZTjrTX8agyM07ETGHlAB3kvQjWi7r74FXI+K1SFev3hP5J3REvEhqo18SYo4ndbf8BvB0pJreixFxXURcV7KMkLrxrpi7ripyjTfSxXWvlow5I5+PgNSVbzxAPtldqrmjopgREa9HxF8i4mBgDVLPnh1JJ/x7qu1dorPGecth5BO6kToi9Cbm70nXgWwTEStFxMqkMaqeyvt6E3PrNsYsrupvFi+FawCPkppj3tiyjCUNDVEm5tXkEUmbtg0hnaR6rWTMG5k/SFfzIFgjKd/9sO0x8+PbNjBfjjeH+V1XZzN/kK3lKF+THknq035Pfh3+k2NfR2qWWVJiLrLGTIkLrqimS/RhpP7wJ5Nq0gfl7aOA63vxvt9ZZl+nY/ZkcRv/EkLSqcBpEfG3LvadHRH7log5mlSbfqSLfVtGxN9LxBwWES93sX0VUiKcviTEbImzM7BlRHy9N3G6iT+cdCXnvb2IMYLUp3sIMDdKjFVTZUxJb46Iu3pbppaYa8B/u0SvQOrRcn/0okt07hjwFtJonKXHZWqJ+RdST6EzGq+h0kWVBwI7RMT2S0LMHh3fid/MbNGUJhuaQOrVs2re/ChpiIljI+LJJSFmj47vxG9mVk7ZHnedjrnQMZz4zczKkXR/RIxZ0mO28pW7ZmbdkLSoietFyR53VcTsCSd+M7PurUYa6bS13V2kqRiXlJiFOfGbmXXvj6QLIae17pB07RIUszC38ZuZ1Yyv3DUzqxknfjOzmnHiH4AkRR7fu3H7K5KOblPs0yXt0Y5YiznOnpJmSrqmZftYSS9Kmpr3T1aaxm5x8TaR1KOpISUNknSCpBmSpku6SdLaPXwqPSJpTr5ieYkh6RP5+d+WX4td8/YDVWD6xdb7STolD+dtfcQndweml4HdJR0T8+ee7XOShkRE0cHMDgY+1dUQFqSB3DbNMdcB/iBJi7noZRPS4G09GT99b9JgZBtFmr1qNGlo4gGr9T3Kz/kbwLhI4/svRxr7BtLwAjOAhxYKtKAF7hcRn2xzsa2HXOMfmF4lTYX3pdYdrTV2Sc/lv1tLuk7SRZJmSzpW0sdyjXq6pHWbwmwvaYrSRNEfzI8fLOkHuVZ8m6RPN8X9ax4j/19dlGefHH+GpO/nbd8E3g2cKukH3T3RiJgNHE4a6AtJm0v6Z/5F8A9J6+dRIL8N7C1pmqS9u7pfF+FXBx6O+VP5zW1cSi/p5/k1uF3St5qezxxJx+TjTJE0TtLlku6R9Jmm1+R6SZdKulPSLyQt9L8oab/8+k+T9Mv8Gg/O72HjV8ii3uNftOk9WhV4FmiM8vpcRNybP0PjgbNy+ZaR9M0ce4akSUq6ut+1mj9i6ELvf97+nKTvSrpVaVLzyvu210rVo8B56fxC+icdQRpNciTwFeDovO90YI/m++a/W5OGhF2dNKTtg8C38r7DgB83Pf4yUqXhTcBcYGnS1IZH5vsMI439v3aO+zywdhflXIM049Ao0q/Pq4Hd8r5rSXMPtz5mLGkAruZtK5CGXSY/7yF5fXvg/Lx+IPDTpsd0eb+WuKPzazgNOB7YtGnfSvnv4FzWjfLtOcBn8/pE0miRy+fn+GjTa/0SsE5+/BWN9yQ/fhXSQGOXMH8S8pNIU/29A7ii+bl3Ue52vkeDgcvz+3Qa8KGmfQu8R43XJK//unHfLu53LenLoLv3P5oef1yj3F7as7jGP0BFGtP+THJNuKCbIuLhSCNl3gM0phOcTkq4Db+PNDb7LNLQuhuQJuj4uKRppKGAVyYlHUhTE3Y1cuVmpMky5kVqXjgLeG8PytvQPLfASOBcSTNIifdti3jMYu8XaX7d9UnTIb4OXCVpu7x7L0m3kCYoeRvQ3GbdmAFsOmnKy2cjjd3/stIolJBek9mR5q39LekXTrPtSEn+pvyabkf6opgNrCPpREk7kiau6Upb3qNcvh2BPUjTGE7Uos8XbSPpRqU5JLZl0a99Q3fv/yukvu6Q5lYYu5hY1s9gLDQAAAJySURBVANu4x/Yfkyajaq57ftVchNfbl5Yqmlf89DIrzfdfp0FPyutF38EKfn+T0Rc3rxD0tZU3y6+KWn2IoDvANdExIcljSXVLrtS6H75S/DPpAm3HwV2kzSb9Ctqs4h4UtLppBp1Q/Pr1vqaNl7Hrl7DZiIN2XtEa5kkbUy66vMzwF7kydQXE6/0exSp2j0ZmCzpCtLn6eiWGEuTfpWMj4gH8pfD0pT3n3xcSHPnOle1kWv8A1hE/Js0m8/BTZvnkGqSALtQbmaiPZV6vKxLqoXeSWoO+KykoZDGb5e07GLiTAa2krSKpMHAPqTJQgrLSfuHwIl500hSMxWk5p2GZ0lNLizmfs2xxyn3RslfkhuRJg4ZQUqUT+e25516UuZsc0lr57h7A60nsa8C9pC0aj7+SpLeqNTjZ1BEnA8cSZpEvStteY8krSGp+RibkF4DWPA1bST5x5VOADf3/Gp97Rt6/f5bOf4WHfiOJ0243nAycJGkW0ntwGVq4/eT/mlHAJ+JiJcknUL6OX6LJJHmZN2tuyAR8bCkCaTJ2gVcGnky+MVYV9JUUrJ5FjghIk7P+44DzpB0JHBp02OuASbkZo5jurlfs1WBkyUNy7cnk84TvJSPfwfwANDjCW1Ic/b+FFgvl+2C5p0R8a9ctr/kL4f/AJ8HXgROazoZvNAvgqwt7xGpYvDD/AX4Un7MZ/K+04FfSHoReBfpszUDeCQ/PxZxv8ZzLPv+Wy95yAazDstNK1+JiA9WFP904I8RcV4V8a3/c1OPmVnNuMZvZlYzrvGbmdWME7+ZWc048ZuZ1YwTv5lZzTjxm5nVjBO/mVnN/C+cVa/eQ/ExRwAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "hist_values = [data[name]['gbtr_depart'].count() for name in data.keys()]\n",
    "fig, ax = plt.subplots()\n",
    "hist, bin_edges = np.histogram(hist_values, bins=15)\n",
    "bin_edges = [int(i) for i in bin_edges]\n",
    "ax.bar(np.arange(0, len(bin_edges) - 1), hist, width=0.5)\n",
    "ax.set_xticks(np.arange(0, len(bin_edges)))\n",
    "ax.set_xticklabels(bin_edges[0:-1], rotation='vertical')\n",
    "ax.set_xlabel('Number of Data Samples per Station')\n",
    "ax.set_ylabel('Frequency of Stations')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We can clearly see that there is obviously too much variation in the distribution of data samples across the station. This shows that there is a minor subset of popular stations within this bike share system. Thus, it might be best if efforts were only focused on developing robust models for the popular stations and using a constant inventory model for the less popular stations. Next, let us plot a more granular graph showing the discrepancy between predicted and actual departure/arrival counts for each model type. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 131,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x360 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "fig, ax = plt.subplots(1, 2, figsize=(10, 5))\n",
    "ax[0].set_title('Departures')\n",
    "ax[1].set_title('Arrivals')\n",
    "for name in data.keys():\n",
    "    actual_d = [x.departure_count for x in data[name]['gbtr_depart'].select('departure_count').collect()]\n",
    "    pred_d = [x.prediction for x in data[name]['gbtr_depart'].select('prediction').collect()]\n",
    "    actual_a = [x.arrival_count for x in data[name]['gbtr_arrival'].select('arrival_count').collect()]\n",
    "    pred_a = [x.prediction for x in data[name]['gbtr_arrival'].select('prediction').collect()]\n",
    "    ax[0].scatter(actual_d, pred_d, color='navy', marker='.')\n",
    "    ax[1].scatter(actual_a, pred_a, color='orange', marker='.')\n",
    "ax[0].set_xlabel('Actual Departures')\n",
    "ax[0].set_ylabel('Predictions')\n",
    "ax[0].set_xlim(0, 90)\n",
    "ax[0].set_ylim(0, 90)\n",
    "ax[1].set_xlabel('Actual Arrivals')\n",
    "ax[1].set_xlim(0, 130)\n",
    "ax[1].set_ylim(0, 130)\n",
    "fig.suptitle('GBT')\n",
    "fig.tight_layout()\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 132,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x360 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "fig, ax = plt.subplots(1, 2, figsize=(10, 5))\n",
    "ax[0].set_title('Departures')\n",
    "ax[1].set_title('Arrivals')\n",
    "for name in data.keys():\n",
    "    actual_d = [x.departure_count for x in data[name]['rfr_depart'].select('departure_count').collect()]\n",
    "    pred_d = [x.prediction for x in data[name]['rfr_depart'].select('prediction').collect()]\n",
    "    actual_a = [x.arrival_count for x in data[name]['rfr_arrival'].select('arrival_count').collect()]\n",
    "    pred_a = [x.prediction for x in data[name]['rfr_arrival'].select('prediction').collect()]\n",
    "    ax[0].scatter(actual_d, pred_d, color='navy', marker='.')\n",
    "    ax[1].scatter(actual_a, pred_a, color='orange', marker='.')\n",
    "ax[0].set_xlabel('Actual Departures')\n",
    "ax[0].set_ylabel('Predictions')\n",
    "ax[0].set_xlim(0, 90)\n",
    "ax[0].set_ylim(0, 90)\n",
    "ax[1].set_xlabel('Actual Arrivals')\n",
    "ax[1].set_xlim(0, 130)\n",
    "ax[1].set_ylim(0, 130)\n",
    "fig.suptitle('RF')\n",
    "fig.tight_layout()\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 133,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x360 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "fig, ax = plt.subplots(1, 2, figsize=(10, 5))\n",
    "ax[0].set_title('Departures')\n",
    "ax[1].set_title('Arrivals')\n",
    "for name in data.keys():\n",
    "    actual_d = [x.departure_count for x in data[name]['lr_depart'].select('departure_count').collect()]\n",
    "    pred_d = [x.prediction for x in data[name]['lr_depart'].select('prediction').collect()]\n",
    "    actual_a = [x.arrival_count for x in data[name]['lr_arrival'].select('arrival_count').collect()]\n",
    "    pred_a = [x.prediction for x in data[name]['lr_arrival'].select('prediction').collect()]\n",
    "    ax[0].scatter(actual_d, pred_d, color='navy', marker='.')\n",
    "    ax[1].scatter(actual_a, pred_a, color='orange', marker='.')\n",
    "ax[0].set_xlabel('Actual Departures')\n",
    "ax[0].set_ylabel('Predictions')\n",
    "ax[0].set_xlim(0, 90)\n",
    "ax[0].set_ylim(0, 90)\n",
    "ax[1].set_xlabel('Actual Arrivals')\n",
    "ax[1].set_xlim(0, 130)\n",
    "ax[1].set_ylim(0, 130)\n",
    "fig.suptitle('LR')\n",
    "fig.tight_layout()\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "These graphs show that the prediction models are severely underestimating both departure and arrival counts. Overall, we can see that GBT is clearly the best out of the three prediction models tested. However, the method of applying the same prediction model to all stations regardless of station popularity is clearly not the best technique. "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "___\n",
    "### Future Steps"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "There is clearly a vast room for improvement in this project. The reduction step from the paper was skipped completely, perhaps adding reduction back into the model is crucial to improving model accuracy. In addition, generating inventory intervals was also skipped, and thus the model from this project was not able to be compared to a constant inventory model. There is also future work that can be done in better utilizing the provided data. For instance, we can see that in this particular bike share system in San Francisco, the majority of stations have under 500 data points. Perhaps this type of modelling can only be completed in cities where bike sharing is so popular that most stations have rental and return trips most hours of the day. Or, in cities where bike sharing is not very popular, perhaps a hybrid model can be developed where prediction models are trained for popular systems, but constant inventory models are used for less popular systems to prevent the use of inaccurate prediction models. "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "___\n",
    "### Bibliography"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "\\[1\\] P. Hulot, D. Aloise, and S.D. Jena, \"Towards Station-Level Demand Prediction for Effective Rebalancing in Bike-Sharing Systems,\" In KDD'18: Proceedings of the 24th ACM SIGKDD International Conference on Knowledge Discover \\& Data Mining, 2018, pp. 378-386. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "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": "3.7.6"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}