model_results.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "#!pip install neo4j\n",
    "#!pip install surprise\n",
    "import pandas as pd\n",
    "import numpy as np\n",
    "from neo4j import GraphDatabase\n",
    "import graph_model as gm\n",
    "from sklearn.metrics import mean_squared_error, mean_absolute_error\n",
    "from surprise import Reader\n",
    "from surprise import Dataset\n",
    "from surprise import accuracy\n",
    "from surprise import KNNBaseline\n",
    "from surprise import SVDpp\n",
    "from surprise import CoClustering\n",
    "from surprise.model_selection import cross_validate\n",
    "import matplotlib.pyplot as plt\n",
    "import seaborn as sns\n",
    "\n",
    "np.set_printoptions(suppress=True)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Model Tuning and Testing #"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "This file contains the final cross validation results for all of the reccomendation systems I experimented with. The results of previous hyperparamter searches are in the cv_results folder. The models were tuned used cross validiton on a few states, mostly Pennsylvainia, Ohio and Illinois.  I then used Quebec as the test set, so none of the models were tuned on this data.\n",
    "\n",
    "This first section is grabbing the Quebec data from the server and making sure we have it all."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>userID</th>\n",
       "      <th>rating</th>\n",
       "      <th>itemID</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>pv6iucL4vcT4Bm69nd477g</td>\n",
       "      <td>1.0</td>\n",
       "      <td>Wrpf3g5EUyuWbf-ey9aQ_A</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>1</td>\n",
       "      <td>d-lhIwhk_9cDGSzF6ZQ-SA</td>\n",
       "      <td>1.0</td>\n",
       "      <td>Wrpf3g5EUyuWbf-ey9aQ_A</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>2</td>\n",
       "      <td>JFt7m1ZtnWhX2dQNZGmvQw</td>\n",
       "      <td>3.0</td>\n",
       "      <td>Wrpf3g5EUyuWbf-ey9aQ_A</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>3</td>\n",
       "      <td>8apKnC2iVfWnfB4cIgUBMw</td>\n",
       "      <td>3.0</td>\n",
       "      <td>Wrpf3g5EUyuWbf-ey9aQ_A</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>4</td>\n",
       "      <td>ZRYqMyXtb-kmHIhDrRdI0Q</td>\n",
       "      <td>4.0</td>\n",
       "      <td>5mVE0xEy9NfMdkhliPh8IQ</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                   userID  rating                  itemID\n",
       "0  pv6iucL4vcT4Bm69nd477g     1.0  Wrpf3g5EUyuWbf-ey9aQ_A\n",
       "1  d-lhIwhk_9cDGSzF6ZQ-SA     1.0  Wrpf3g5EUyuWbf-ey9aQ_A\n",
       "2  JFt7m1ZtnWhX2dQNZGmvQw     3.0  Wrpf3g5EUyuWbf-ey9aQ_A\n",
       "3  8apKnC2iVfWnfB4cIgUBMw     3.0  Wrpf3g5EUyuWbf-ey9aQ_A\n",
       "4  ZRYqMyXtb-kmHIhDrRdI0Q     4.0  5mVE0xEy9NfMdkhliPh8IQ"
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "uri = \"bolt://54.145.141.152:7687\"''\n",
    "driver = GraphDatabase.driver(uri, auth=(\"neo4j\", \"i-0e23d19f0d8795714\"))\n",
    "\n",
    "result = gm.cypher(driver,'\\\n",
    "    MATCH (u:User)-->(r:Review)-->(b:Business)-->(:City)-->(s:State {name:\"QC\"})\\\n",
    "    RETURN u.id, r.stars, b.id', [\"userID\", \"rating\", \"itemID\"])\n",
    "\n",
    "QC_reviewsdf = result\n",
    "QC_reviewsdf.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "QC_reviewsdf.to_csv('Q')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(178971, 3)"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "QC_reviewsdf.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "userID    52738\n",
       "rating        5\n",
       "itemID     9210\n",
       "dtype: int64"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "QC_reviewsdf.nunique()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "#Now we put the data into a dataset object for use by the surprise package testing methods\n",
    "reader = Reader(rating_scale=(1, 5))\n",
    "data = Dataset.load_from_df(QC_reviewsdf[['userID', 'itemID', 'rating']], reader)\n",
    "\n",
    "# for the graph model we have to take a random sample of businesses to test. The rest of the businesses are potentially\n",
    "#used in the calculating predicted reviews.\n",
    "test_sample = QC_reviewsdf.sample(4000)\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Testing the Graph Model ##"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [],
   "source": [
    "test_sample['prediction'] = [gm.graph_model_predict(driver, x , y) for x,y in zip(test_sample['userID'], test_sample['itemID'])]\n",
    "test_sample['error'] = test_sample.rating-test_sample.prediction\n",
    "\n",
    "rms = pow(mean_squared_error(test_sample.rating, test_sample.prediction), .5)\n",
    "mae = mean_absolute_error(test_sample.rating, test_sample.prediction)\n",
    "rms,mae"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.axes._subplots.AxesSubplot at 0x7fec1c3216a0>"
      ]
     },
     "execution_count": 20,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "sns.residplot(test_sample.rating, test_sample.prediction, color=\"g\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Evaluation of prediciton models: SVDpp, Coclustering, and KNN with baseline esitamates both user and item focused. We also investigate the effectiveness of ensembling these methods and taking an mean and median average. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "coclu=CoClustering(n_cltr_u=2, n_cltr_i=6, n_epochs=40)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "svd=SVDpp(n_factors=15, n_epochs=15, lr_all=.02, reg_all=.2)\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "sim_options = {'name': 'pearson_baseline',\n",
    "               'user_based': False,\n",
    "               'min_support': 10,\n",
    "               'shrinkage': 50\n",
    "              }\n",
    "\n",
    "bsl_options = {'method': 'als',\n",
    "               'n_epochs': 15,\n",
    "               'reg_u': 5,\n",
    "               'reg_i': 2\n",
    "               }\n",
    "\n",
    "\n",
    "knn=KNNBaseline(sim_options=sim_options, bsl_options=bsl_options, k=30, min_k=15)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Evaluating RMSE, MAE of algorithm CoClustering on 5 split(s).\n",
      "\n",
      "                  Fold 1  Fold 2  Fold 3  Fold 4  Fold 5  Mean    Std     \n",
      "RMSE (testset)    1.2510  1.2553  1.2529  1.2613  1.2627  1.2566  0.0046  \n",
      "MAE (testset)     0.9402  0.9403  0.9395  0.9456  0.9488  0.9429  0.0037  \n",
      "Fit time          14.19   13.97   14.27   14.21   14.23   14.17   0.10    \n",
      "Test time         0.32    0.32    0.20    0.32    0.33    0.30    0.05    \n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "{'test_rmse': array([1.2510053 , 1.25525689, 1.25287729, 1.2613098 , 1.26271075]),\n",
       " 'test_mae': array([0.94023614, 0.94025611, 0.93951462, 0.9455891 , 0.94877878]),\n",
       " 'fit_time': (14.18798279762268,\n",
       "  13.974129676818848,\n",
       "  14.26918888092041,\n",
       "  14.20827841758728,\n",
       "  14.228464365005493),\n",
       " 'test_time': (0.3239619731903076,\n",
       "  0.3242943286895752,\n",
       "  0.20160865783691406,\n",
       "  0.32163500785827637,\n",
       "  0.32757997512817383)}"
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "cross_validate(coclu, data, verbose=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Evaluating RMSE, MAE of algorithm SVDpp on 5 split(s).\n",
      "\n",
      "                  Fold 1  Fold 2  Fold 3  Fold 4  Fold 5  Mean    Std     \n",
      "RMSE (testset)    1.1241  1.1286  1.1265  1.1251  1.1374  1.1284  0.0048  \n",
      "MAE (testset)     0.8790  0.8812  0.8822  0.8803  0.8886  0.8823  0.0033  \n",
      "Fit time          44.12   47.11   44.53   44.78   44.97   45.10   1.05    \n",
      "Test time         1.89    2.03    2.07    2.09    2.00    2.01    0.07    \n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "{'test_rmse': array([1.12414642, 1.12860151, 1.12652921, 1.12507021, 1.13744645]),\n",
       " 'test_mae': array([0.87896629, 0.88122054, 0.88224494, 0.88031921, 0.88857595]),\n",
       " 'fit_time': (44.12103247642517,\n",
       "  47.1144483089447,\n",
       "  44.52857279777527,\n",
       "  44.77706813812256,\n",
       "  44.97261071205139),\n",
       " 'test_time': (1.8885676860809326,\n",
       "  2.0316741466522217,\n",
       "  2.066457509994507,\n",
       "  2.0894460678100586,\n",
       "  1.9984714984893799)}"
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "cross_validate(svd, data, verbose=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Estimating biases using als...\n",
      "Computing the pearson_baseline similarity matrix...\n",
      "Done computing similarity matrix.\n",
      "Estimating biases using als...\n",
      "Computing the pearson_baseline similarity matrix...\n",
      "Done computing similarity matrix.\n",
      "Estimating biases using als...\n",
      "Computing the pearson_baseline similarity matrix...\n",
      "Done computing similarity matrix.\n",
      "Estimating biases using als...\n",
      "Computing the pearson_baseline similarity matrix...\n",
      "Done computing similarity matrix.\n",
      "Estimating biases using als...\n",
      "Computing the pearson_baseline similarity matrix...\n",
      "Done computing similarity matrix.\n",
      "Evaluating RMSE, MAE of algorithm KNNBaseline on 5 split(s).\n",
      "\n",
      "                  Fold 1  Fold 2  Fold 3  Fold 4  Fold 5  Mean    Std     \n",
      "RMSE (testset)    1.1147  1.1262  1.1238  1.1237  1.1212  1.1219  0.0040  \n",
      "MAE (testset)     0.8706  0.8790  0.8730  0.8765  0.8733  0.8745  0.0030  \n",
      "Fit time          5.92    6.00    5.99    6.05    5.98    5.99    0.04    \n",
      "Test time         1.42    1.40    1.62    1.61    1.62    1.53    0.10    \n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "{'test_rmse': array([1.1146612 , 1.12624527, 1.12378043, 1.12366525, 1.12120049]),\n",
       " 'test_mae': array([0.87055734, 0.87903172, 0.87297376, 0.87646662, 0.87328342]),\n",
       " 'fit_time': (5.918645143508911,\n",
       "  5.99767804145813,\n",
       "  5.991095781326294,\n",
       "  6.049122333526611,\n",
       "  5.979476451873779),\n",
       " 'test_time': (1.4174723625183105,\n",
       "  1.3976738452911377,\n",
       "  1.6226787567138672,\n",
       "  1.6135437488555908,\n",
       "  1.6180102825164795)}"
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "cross_validate(knn, data, verbose=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "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": "3.7.4"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}