03-User-Based-Collaborative-Filtering.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 155,
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "import seaborn as sns\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "\n",
    "from pyspark.sql import SparkSession\n",
    "\n",
    "from pyspark.sql import Window\n",
    "from pyspark.sql.functions import countDistinct, collect_list, collect_set, create_map, col, mean, avg, row_number, desc, udf, count, lit, array_intersect, size, rank, concat, array, sum, abs, when\n"
   ],
   "metadata": {
    "collapsed": false
   }
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "outputs": [],
   "source": [
    "spark = SparkSession.builder.master(\"local[4]\") \\\n",
    "                    .appName('recommendation_system') \\\n",
    "                    .config(\"spark.driver.memory\", \"15g\") \\\n",
    "                    .getOrCreate()"
   ],
   "metadata": {
    "collapsed": false
   }
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "outputs": [],
   "source": [
    "df = spark.read.options(\n",
    "            header=True,\n",
    "            inferSchema=True\n",
    "        ) \\\n",
    "        .csv('data/ratings_processed.csv')"
   ],
   "metadata": {
    "collapsed": false
   }
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "collapsed": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "The unique number of users is 2000\n",
      "User ID starts from 0\n",
      "User ID ends at 1999\n"
     ]
    }
   ],
   "source": [
    "print(f\"The unique number of users is {df.select(countDistinct('userId')).collect()[0][0]}\\n\"\n",
    "      f\"User ID starts from {df.agg({'userId': 'min'}).collect()[0][0]}\\n\"\n",
    "      f\"User ID ends at {df.agg({'userId': 'max'}).collect()[0][0]}\"\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "The unique number of movies is 500\n",
      "Movie ID starts from 0\n",
      "Movie ID ends at 499\n"
     ]
    }
   ],
   "source": [
    "print(f\"The unique number of movies is {df.select(countDistinct('movieId')).collect()[0][0]}\\n\"\n",
    "      f\"Movie ID starts from {df.agg({'movieId': 'min'}).collect()[0][0]}\\n\"\n",
    "      f\"Movie ID ends at {df.agg({'movieId': 'max'}).collect()[0][0]}\"\n",
    ")"
   ],
   "metadata": {
    "collapsed": false
   }
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "outputs": [],
   "source": [
    "N = df.agg({'userId': 'max'}).collect()[0][0]\n",
    "M = df.agg({'movieId': 'max'}).collect()[0][0]"
   ],
   "metadata": {
    "collapsed": false
   }
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "outputs": [],
   "source": [
    "train_data, test_data = df.randomSplit([0.8, 0.2], seed=42)"
   ],
   "metadata": {
    "collapsed": false
   }
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "The number of samples in training data is 542114\n",
      "The number of samples in test data is 135415\n"
     ]
    }
   ],
   "source": [
    "print(\n",
    "    f\"The number of samples in training data is {train_data.count()}\\n\"\n",
    "    f\"The number of samples in test data is {test_data.count()}\"\n",
    ")"
   ],
   "metadata": {
    "collapsed": false
   }
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "root\n",
      " |-- rating: double (nullable = true)\n",
      " |-- userId: integer (nullable = true)\n",
      " |-- movieId: integer (nullable = true)\n",
      "\n"
     ]
    }
   ],
   "source": [
    "train_data.printSchema()"
   ],
   "metadata": {
    "collapsed": false
   }
  },
  {
   "cell_type": "markdown",
   "source": [
    "### Checking if the entries of all users and movies are present in both train and test set"
   ],
   "metadata": {
    "collapsed": false
   }
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "The number of users in training set is 2000\n",
      "The number of users in test set is 2000\n",
      "The number of movies in training set is 500\n",
      "The number of movies in test set is 500\n",
      "\n"
     ]
    }
   ],
   "source": [
    "print(\n",
    "    f\"The number of users in training set is {train_data.select(countDistinct('userId')).collect()[0][0]}\\n\"\n",
    "    f\"The number of users in test set is {test_data.select(countDistinct('userId')).collect()[0][0]}\\n\"\n",
    "    f\"The number of movies in training set is {train_data.select(countDistinct('movieId')).collect()[0][0]}\\n\"\n",
    "    f\"The number of movies in test set is {test_data.select(countDistinct('movieId')).collect()[0][0]}\\n\"\n",
    ")"
   ],
   "metadata": {
    "collapsed": false
   }
  },
  {
   "cell_type": "markdown",
   "source": [
    "### Distribution of movies in training and test data"
   ],
   "metadata": {
    "collapsed": false
   }
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "outputs": [
    {
     "data": {
      "text/plain": "<Figure size 1440x576 with 2 Axes>",
      "image/png": "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\n"
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "fig, axs = plt.subplots(1, 2, figsize=(20, 8))\n",
    "sns.histplot(\n",
    "    data = train_data.toPandas(),\n",
    "    x = 'movieId',\n",
    "    element='poly',\n",
    "    ax =axs[0]\n",
    ")\n",
    "\n",
    "axs[0].set_title('Distribution of movies in training set')\n",
    "sns.histplot(\n",
    "    data = test_data.toPandas(),\n",
    "    x = 'movieId',\n",
    "    element='poly',\n",
    "    ax =axs[1]\n",
    ")\n",
    "axs[1].set_title('Distribution of movies in test set')\n",
    "plt.show()"
   ],
   "metadata": {
    "collapsed": false
   }
  },
  {
   "cell_type": "code",
   "execution_count": 219,
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "root\n",
      " |-- rating: double (nullable = true)\n",
      " |-- userId: integer (nullable = true)\n",
      " |-- movieId: integer (nullable = true)\n",
      "\n"
     ]
    }
   ],
   "source": [
    "train_data.printSchema()"
   ],
   "metadata": {
    "collapsed": false
   }
  },
  {
   "cell_type": "code",
   "execution_count": 220,
   "outputs": [],
   "source": [
    "## Find sigma value\n",
    "def find_sigma(dev_list):\n",
    "    arr = np.array(dev_list)\n",
    "    result = np.sqrt(arr.dot(arr))\n",
    "    return int(result)"
   ],
   "metadata": {
    "collapsed": false
   }
  },
  {
   "cell_type": "code",
   "execution_count": 221,
   "outputs": [
    {
     "ename": "AssertionError",
     "evalue": "",
     "output_type": "error",
     "traceback": [
      "\u001B[0;31m---------------------------------------------------------------------------\u001B[0m",
      "\u001B[0;31mAssertionError\u001B[0m                            Traceback (most recent call last)",
      "Input \u001B[0;32mIn [221]\u001B[0m, in \u001B[0;36m<cell line: 4>\u001B[0;34m()\u001B[0m\n\u001B[1;32m      1\u001B[0m \u001B[38;5;28;01mfrom\u001B[39;00m \u001B[38;5;21;01mpyspark\u001B[39;00m\u001B[38;5;21;01m.\u001B[39;00m\u001B[38;5;21;01msql\u001B[39;00m \u001B[38;5;28;01mimport\u001B[39;00m types \u001B[38;5;28;01mas\u001B[39;00m T\n\u001B[1;32m      2\u001B[0m find_sigma_udf \u001B[38;5;241m=\u001B[39m udf(find_sigma, T\u001B[38;5;241m.\u001B[39mIntegerType())\n\u001B[0;32m----> 4\u001B[0m window \u001B[38;5;241m=\u001B[39m \u001B[43mWindow\u001B[49m\u001B[38;5;241;43m.\u001B[39;49m\u001B[43mpartitionBy\u001B[49m\u001B[43m(\u001B[49m\u001B[38;5;124;43m'\u001B[39;49m\u001B[38;5;124;43muserId\u001B[39;49m\u001B[38;5;124;43m'\u001B[39;49m\u001B[43m)\u001B[49m\u001B[38;5;241m.\u001B[39morderBy(\u001B[38;5;124m'\u001B[39m\u001B[38;5;124mrating\u001B[39m\u001B[38;5;124m'\u001B[39m)\n\u001B[1;32m      5\u001B[0m windowAgg \u001B[38;5;241m=\u001B[39m Window\u001B[38;5;241m.\u001B[39mpartitionBy(\u001B[38;5;124m'\u001B[39m\u001B[38;5;124muserId\u001B[39m\u001B[38;5;124m'\u001B[39m)\n\u001B[1;32m      7\u001B[0m summarised_train_data \u001B[38;5;241m=\u001B[39m train_data \\\n\u001B[1;32m      8\u001B[0m     \u001B[38;5;241m.\u001B[39mwithColumn(\u001B[38;5;124m\"\u001B[39m\u001B[38;5;124mrow\u001B[39m\u001B[38;5;124m\"\u001B[39m, row_number()\u001B[38;5;241m.\u001B[39mover(window)) \\\n\u001B[1;32m      9\u001B[0m     \u001B[38;5;241m.\u001B[39mwithColumn(\u001B[38;5;124m\"\u001B[39m\u001B[38;5;124mavg\u001B[39m\u001B[38;5;124m\"\u001B[39m, avg(col(\u001B[38;5;124m\"\u001B[39m\u001B[38;5;124mrating\u001B[39m\u001B[38;5;124m\"\u001B[39m))\u001B[38;5;241m.\u001B[39mover(windowAgg)) \\\n\u001B[0;32m   (...)\u001B[0m\n\u001B[1;32m     23\u001B[0m     \u001B[38;5;241m.\u001B[39mtoDF([\u001B[38;5;124m\"\u001B[39m\u001B[38;5;124muserId\u001B[39m\u001B[38;5;124m\"\u001B[39m, \u001B[38;5;124m\"\u001B[39m\u001B[38;5;124mmovies\u001B[39m\u001B[38;5;124m\"\u001B[39m, \u001B[38;5;124m\"\u001B[39m\u001B[38;5;124mcalculations\u001B[39m\u001B[38;5;124m\"\u001B[39m]) \\\n\u001B[1;32m     24\u001B[0m     \u001B[38;5;241m.\u001B[39morderBy(\u001B[38;5;124m'\u001B[39m\u001B[38;5;124muserId\u001B[39m\u001B[38;5;124m'\u001B[39m)\n",
      "File \u001B[0;32m~/miniconda/envs/personal/lib/python3.9/site-packages/pyspark/sql/window.py:79\u001B[0m, in \u001B[0;36mWindow.partitionBy\u001B[0;34m(*cols)\u001B[0m\n\u001B[1;32m     75\u001B[0m \u001B[38;5;124;03m\"\"\"\u001B[39;00m\n\u001B[1;32m     76\u001B[0m \u001B[38;5;124;03mCreates a :class:`WindowSpec` with the partitioning defined.\u001B[39;00m\n\u001B[1;32m     77\u001B[0m \u001B[38;5;124;03m\"\"\"\u001B[39;00m\n\u001B[1;32m     78\u001B[0m sc \u001B[38;5;241m=\u001B[39m SparkContext\u001B[38;5;241m.\u001B[39m_active_spark_context\n\u001B[0;32m---> 79\u001B[0m \u001B[38;5;28;01massert\u001B[39;00m sc \u001B[38;5;129;01mis\u001B[39;00m \u001B[38;5;129;01mnot\u001B[39;00m \u001B[38;5;28;01mNone\u001B[39;00m \u001B[38;5;129;01mand\u001B[39;00m sc\u001B[38;5;241m.\u001B[39m_jvm \u001B[38;5;129;01mis\u001B[39;00m \u001B[38;5;129;01mnot\u001B[39;00m \u001B[38;5;28;01mNone\u001B[39;00m\n\u001B[1;32m     80\u001B[0m jspec \u001B[38;5;241m=\u001B[39m sc\u001B[38;5;241m.\u001B[39m_jvm\u001B[38;5;241m.\u001B[39morg\u001B[38;5;241m.\u001B[39mapache\u001B[38;5;241m.\u001B[39mspark\u001B[38;5;241m.\u001B[39msql\u001B[38;5;241m.\u001B[39mexpressions\u001B[38;5;241m.\u001B[39mWindow\u001B[38;5;241m.\u001B[39mpartitionBy(_to_java_cols(cols))\n\u001B[1;32m     81\u001B[0m \u001B[38;5;28;01mreturn\u001B[39;00m WindowSpec(jspec)\n",
      "\u001B[0;31mAssertionError\u001B[0m: "
     ]
    }
   ],
   "source": [
    "from pyspark.sql import types as T\n",
    "find_sigma_udf = udf(find_sigma, T.IntegerType())\n",
    "\n",
    "window = Window.partitionBy('userId').orderBy('rating')\n",
    "windowAgg = Window.partitionBy('userId')\n",
    "\n",
    "summarised_train_data = train_data \\\n",
    "    .withColumn(\"row\", row_number().over(window)) \\\n",
    "    .withColumn(\"avg\", avg(col(\"rating\")).over(windowAgg)) \\\n",
    "    .withColumn(\"deviation\", col('rating') - col('avg')) \\\n",
    "    .withColumn(\"sigma\", find_sigma_udf(collect_list(col('deviation')).over(windowAgg))) \\\n",
    "    .withColumn(\"calculations\", create_map(\n",
    "        lit(\"avg\"), col(\"avg\"),\n",
    "        lit(\"sigma\"), col(\"sigma\"),\n",
    "        col(\"movieId\"), col(\"deviation\"),\n",
    "        )) \\\n",
    "    .withColumn(\"movies\", collect_set(col('movieId')).over(windowAgg)) \\\n",
    "    .select('userId', 'movies', 'calculations') \\\n",
    "    .rdd \\\n",
    "    .map(lambda row: ((row[0], tuple(row[1])), row[2])) \\\n",
    "    .reduceByKey(lambda a, b: {**a, **b}) \\\n",
    "    .map(lambda row: (row[0][0], list(row[0][1]), row[1])) \\\n",
    "    .toDF([\"userId\", \"movies\", \"calculations\"]) \\\n",
    "    .orderBy('userId')\n"
   ],
   "metadata": {
    "collapsed": false
   }
  },
  {
   "cell_type": "code",
   "execution_count": 115,
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "+------+--------------------+--------------------+\n",
      "|userId|              movies|        calculations|\n",
      "+------+--------------------+--------------------+\n",
      "|     0|[364, 356, 212, 1...|{191 -> 0.8963414...|\n",
      "|     1|[0, 356, 306, 387...|{471 -> -0.161290...|\n",
      "|     2|[0, 387, 277, 256...|{350 -> 0.8489208...|\n",
      "|     3|[0, 356, 466, 306...|{230 -> 0.8745019...|\n",
      "|     4|[0, 306, 387, 277...|{350 -> 0.5394190...|\n",
      "|     5|[466, 356, 387, 2...|{350 -> -0.295275...|\n",
      "|     6|[466, 356, 306, 1...|{471 -> 0.5919282...|\n",
      "|     7|[466, 0, 306, 2, ...|{471 -> 0.1887254...|\n",
      "|     8|[306, 387, 277, 2...|{350 -> 0.0540069...|\n",
      "|     9|[0, 356, 466, 437...|{471 -> 0.0988593...|\n",
      "|    10|[466, 356, 212, 3...|{230 -> 0.765625,...|\n",
      "|    11|[0, 356, 387, 256...|{350 -> 0.65625, ...|\n",
      "|    12|[306, 387, 227, 2...|{470 -> -0.100378...|\n",
      "|    13|[466, 437, 306, 3...|{470 -> -0.777173...|\n",
      "|    14|[0, 466, 356, 387...|{470 -> 1.3665480...|\n",
      "|    15|[356, 306, 277, 2...|{350 -> -0.314199...|\n",
      "|    16|[0, 466, 102, 306...|{350 -> 0.4200913...|\n",
      "|    17|[306, 277, 256, 2...|{230 -> 0.4244966...|\n",
      "|    18|[0, 466, 437, 387...|{470 -> -0.268115...|\n",
      "|    19|[0, 466, 437, 306...|{470 -> 0.1333333...|\n",
      "+------+--------------------+--------------------+\n",
      "only showing top 20 rows\n",
      "\n"
     ]
    }
   ],
   "source": [
    "summarised_train_data.show()"
   ],
   "metadata": {
    "collapsed": false
   }
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "outputs": [
    {
     "data": {
      "text/plain": "2000"
     },
     "execution_count": 19,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "summarised_train_data.count()"
   ],
   "metadata": {
    "collapsed": false
   }
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "outputs": [],
   "source": [
    "K = 25\n",
    "limit = 5"
   ],
   "metadata": {
    "collapsed": false
   }
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "outputs": [],
   "source": [
    "def calculate_weight(calculations_i, calculations_j, common_movies):\n",
    "    numerator = 0\n",
    "    for movieId in common_movies:\n",
    "        numerator += calculations_i[str(movieId)] * calculations_j[str(movieId)]\n",
    "\n",
    "    return numerator / ( calculations_i['sigma'] * calculations_j['sigma'] )"
   ],
   "metadata": {
    "collapsed": false
   }
  },
  {
   "cell_type": "code",
   "execution_count": 99,
   "outputs": [],
   "source": [
    "\n",
    "windowAgg = Window.partitionBy(\"user_i\").orderBy(desc(\"weight\"))\n",
    "\n",
    "calculate_weight_udf = udf(calculate_weight, T.FloatType())\n",
    "\n",
    "weight_data = summarised_train_data \\\n",
    "    .select(\n",
    "        col('userId').alias('user_i'),\n",
    "        col('movies').alias('movies_i'),\n",
    "        col('calculations').alias('calculations_i')\n",
    "    ) \\\n",
    "    .crossJoin(summarised_train_data\n",
    "               .select(\n",
    "                        col('userId').alias('user_j'),\n",
    "                        col('movies').alias('movies_j'),\n",
    "                        col('calculations').alias('calculations_j'),\n",
    "                    )\n",
    "               ) \\\n",
    "    .filter(col('user_i') != col('user_j')) \\\n",
    "    .withColumn(\n",
    "                \"common_movies\", array_intersect(col('movies_i'), col('movies_j'))\n",
    "                ) \\\n",
    "    .withColumn(\n",
    "                \"weight\", calculate_weight_udf(\n",
    "                                    col(\"calculations_i\"),\n",
    "                                    col(\"calculations_j\"),\n",
    "                                    col(\"common_movies\"),\n",
    "                                )\n",
    "    ) \\\n",
    "    .drop(\"movies_i\", \"movies_j\", \"calculations_i\", \"calculations_j\", \"common_movies\") \\\n",
    "    .withColumn(\"weight_ij\", create_map(\n",
    "        col('user_j'), col(\"weight\"),\n",
    "        )) \\\n",
    "    .withColumn(\"row\", row_number().over(windowAgg)) \\\n",
    "    .orderBy(col(\"user_i\"), col(\"weight\").desc()) \\\n",
    "    .drop(\"user_j\") \\\n",
    "    .filter(col(\"row\") <= K).drop(\"row\")\n",
    "    # .rdd \\\n",
    "    # .map(lambda row: (row[0], row[2])).reduceByKey(lambda a, b: {**a, **b}) \\\n",
    "    # .map(lambda row:\n",
    "    #          (\n",
    "    #              row[0],\n",
    "    #               [\n",
    "    #                   {k: float(v) for k, v in [('neighbor', neighbor), ('weight', weight)]}\n",
    "    #                   for neighbor, weight in sorted(row[1].items(), key=lambda item: item[1], reverse=True)\n",
    "    #               ]\n",
    "    #          )\n",
    "    #      ) \\\n",
    "    # .toDF([\"user\", \"neighbors_weight\"])\n"
   ],
   "metadata": {
    "collapsed": false
   }
  },
  {
   "cell_type": "code",
   "execution_count": 108,
   "outputs": [],
   "source": [
    "\n",
    "windowAgg = Window.partitionBy(\"user_i\").orderBy(desc(\"weight\"))\n",
    "\n",
    "calculate_weight_udf = udf(calculate_weight, T.FloatType())\n",
    "\n",
    "weight_data = summarised_train_data \\\n",
    "    .select(\n",
    "        col('userId').alias('user_i'),\n",
    "        col('movies').alias('movies_i'),\n",
    "        col('calculations').alias('calculations_i')\n",
    "    ) \\\n",
    "    .crossJoin(summarised_train_data\n",
    "               .select(\n",
    "                        col('userId').alias('user_j'),\n",
    "                        col('movies').alias('movies_j'),\n",
    "                        col('calculations').alias('calculations_j'),\n",
    "                    )\n",
    "               ) \\\n",
    "    .filter(col('user_i') != col('user_j')) \\\n",
    "    .withColumn(\n",
    "                \"common_movies\", array_intersect(col('movies_i'), col('movies_j'))\n",
    "                ) \\\n",
    "    .withColumn(\n",
    "                \"weight\", calculate_weight_udf(\n",
    "                                    col(\"calculations_i\"),\n",
    "                                    col(\"calculations_j\"),\n",
    "                                    col(\"common_movies\"),\n",
    "                                )\n",
    "    ) \\\n",
    "    .drop(\"movies_i\", \"movies_j\", \"common_movies\") \\\n",
    "    .withColumn(\"row\", row_number().over(windowAgg)) \\\n",
    "    .filter(col(\"row\") <= K).drop(\"row\")\n",
    "    # .rdd \\\n",
    "    # .map(lambda row: (row[0], row[2])).reduceByKey(lambda a, b: {**a, **b}) \\\n",
    "    # .map(lambda row:\n",
    "    #          (\n",
    "    #              row[0],\n",
    "    #               [\n",
    "    #                   {k: float(v) for k, v in [('neighbor', neighbor), ('weight', weight)]}\n",
    "    #                   for neighbor, weight in sorted(row[1].items(), key=lambda item: item[1], reverse=True)\n",
    "    #               ]\n",
    "    #          )\n",
    "    #      ) \\\n",
    "    # .toDF([\"user\", \"neighbors_weight\"])\n"
   ],
   "metadata": {
    "collapsed": false
   }
  },
  {
   "cell_type": "code",
   "execution_count": 109,
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[Stage 25740:=================================================>   (15 + 1) / 16]\r"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "+------+--------------------+------+--------------------+----------+\n",
      "|user_i|      calculations_i|user_j|      calculations_j|    weight|\n",
      "+------+--------------------+------+--------------------+----------+\n",
      "|    26|{470 -> -0.052478...|  1989|{470 -> 0.0944148...|0.44695708|\n",
      "|    26|{470 -> -0.052478...|  1013|{350 -> 0.1464435...| 0.4464242|\n",
      "|    26|{470 -> -0.052478...|  1451|{471 -> 1.6782178...|0.44227955|\n",
      "|    26|{470 -> -0.052478...|  1306|{471 -> -0.263297...|0.43343735|\n",
      "|    26|{470 -> -0.052478...|  1081|{470 -> 0.7353801...|0.43270737|\n",
      "|    26|{470 -> -0.052478...|  1300|{0 -> 0.779448621...|0.43129292|\n",
      "|    26|{470 -> -0.052478...|   973|{470 -> 0.1521164...|0.42491606|\n",
      "|    26|{470 -> -0.052478...|  1191|{350 -> -0.886435...|0.41981676|\n",
      "|    26|{470 -> -0.052478...|   827|{470 -> -0.373437...|0.41726848|\n",
      "|    26|{470 -> -0.052478...|  1571|{470 -> 0.6873111...|0.41580778|\n",
      "|    26|{470 -> -0.052478...|   322|{350 -> -0.496282...|  0.414136|\n",
      "|    26|{470 -> -0.052478...|  1855|{470 -> 0.0668789...|0.40740103|\n",
      "|    26|{470 -> -0.052478...|   845|{471 -> -0.454682...|0.40646857|\n",
      "|    26|{470 -> -0.052478...|  1528|{350 -> -0.767657...|0.40552193|\n",
      "|    26|{470 -> -0.052478...|   114|{1 -> -0.83924050...| 0.4032979|\n",
      "|    26|{470 -> -0.052478...|   498|{470 -> 0.1912350...|0.40261686|\n",
      "|    26|{470 -> -0.052478...|   251|{470 -> 0.0934959...|0.40018857|\n",
      "|    26|{470 -> -0.052478...|   963|{350 -> -0.437919...| 0.3992827|\n",
      "|    26|{470 -> -0.052478...|  1367|{470 -> 0.1666666...| 0.3988018|\n",
      "|    26|{470 -> -0.052478...|    19|{470 -> 0.1333333...| 0.3977561|\n",
      "+------+--------------------+------+--------------------+----------+\n",
      "only showing top 20 rows\n",
      "\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "                                                                                \r"
     ]
    }
   ],
   "source": [
    "weight_data.show()"
   ],
   "metadata": {
    "collapsed": false
   }
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "outputs": [],
   "source": [
    "weight_data.write.json('./data/user_weights.json')"
   ],
   "metadata": {
    "collapsed": false
   }
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "outputs": [],
   "source": [
    "weight_data = spark.read.json('./data/user_weights.json')\n",
    "weight_data = weight_data.select('user', 'neighbors_weight')"
   ],
   "metadata": {
    "collapsed": false
   }
  },
  {
   "cell_type": "markdown",
   "source": [
    "### Evaluating Training Dataset"
   ],
   "metadata": {
    "collapsed": false
   }
  },
  {
   "cell_type": "code",
   "execution_count": 150,
   "outputs": [
    {
     "data": {
      "text/plain": "542114"
     },
     "execution_count": 150,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train_data.count()"
   ],
   "metadata": {
    "collapsed": false
   }
  },
  {
   "cell_type": "code",
   "execution_count": 203,
   "outputs": [],
   "source": [
    "window = Window.partitionBy('user_i', 'movieId')\n",
    "windowAgg = Window.partitionBy('user_i', 'movieId').orderBy(\"userId\", \"movieId\")\n",
    "\n",
    "train_predictions = train_data.join(weight_data,train_data.userId == weight_data.user_i, \"inner\").drop('user_Id') \\\n",
    "    .withColumn(\"weighted_deviation\", col('weight') * col('calculations_j')[col('movieId')]) \\\n",
    "    .filter(col('weighted_deviation').isNotNull()) \\\n",
    "    .withColumn(\"numerator\", sum(col('weighted_deviation')).over(window) ) \\\n",
    "    .withColumn(\"denominator\", sum(abs(col('weight'))).over(window) ) \\\n",
    "    .withColumn(\"prediction\", col(\"numerator\") / col(\"denominator\") + col('calculations_i')['avg']) \\\n",
    "    .withColumn(\"row\", row_number().over(windowAgg)) \\\n",
    "    .filter(col(\"row\") == 1).drop(\"row\") \\\n",
    "    .select(\"userId\", \"movieId\", \"rating\", \"prediction\")"
   ],
   "metadata": {
    "collapsed": false
   }
  },
  {
   "cell_type": "code",
   "execution_count": 204,
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[Stage 26822:=================================================>   (15 + 1) / 16]\r"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "+------+-------+------+------------------+\n",
      "|userId|movieId|rating|        prediction|\n",
      "+------+-------+------+------------------+\n",
      "|    26|      1|   3.0|2.8028858092072486|\n",
      "|    26|      3|   3.0|2.2921321351116397|\n",
      "|    26|      5|   4.0| 2.906995424047466|\n",
      "|    26|      7|   4.0|3.6171285312119656|\n",
      "|    26|      9|   4.0|3.7289351139936238|\n",
      "|    26|     10|   2.0|1.6293744869676965|\n",
      "|    26|     11|   3.0| 3.548868593404618|\n",
      "|    26|     12|   3.5| 3.075381112122394|\n",
      "|    26|     14|   3.0|2.6019674150897845|\n",
      "|    26|     15|   3.5| 4.019432460521232|\n",
      "+------+-------+------+------------------+\n",
      "only showing top 10 rows\n",
      "\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "                                                                                \r"
     ]
    }
   ],
   "source": [
    "train_predictions.show(10)"
   ],
   "metadata": {
    "collapsed": false
   }
  },
  {
   "cell_type": "code",
   "execution_count": 208,
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "                                                                                \r"
     ]
    },
    {
     "data": {
      "text/plain": "0.5198600159527748"
     },
     "execution_count": 208,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train_predictions \\\n",
    "    .withColumn('error', (col('prediction') - col('rating')) * (col('prediction') - col('rating'))) \\\n",
    "    .select(mean(col('error'))) \\\n",
    "    .collect()[0][0]"
   ],
   "metadata": {
    "collapsed": false
   }
  },
  {
   "cell_type": "markdown",
   "source": [
    "### Evaluating Test Dataset"
   ],
   "metadata": {
    "collapsed": false
   }
  },
  {
   "cell_type": "code",
   "execution_count": 209,
   "outputs": [
    {
     "data": {
      "text/plain": "135415"
     },
     "execution_count": 209,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "test_data.count()"
   ],
   "metadata": {
    "collapsed": false
   }
  },
  {
   "cell_type": "code",
   "execution_count": 210,
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "root\n",
      " |-- rating: double (nullable = true)\n",
      " |-- userId: integer (nullable = true)\n",
      " |-- movieId: integer (nullable = true)\n",
      "\n"
     ]
    }
   ],
   "source": [
    "test_data.printSchema()"
   ],
   "metadata": {
    "collapsed": false
   }
  },
  {
   "cell_type": "code",
   "execution_count": 212,
   "outputs": [],
   "source": [
    "window = Window.partitionBy('user_i', 'movieId')\n",
    "windowAgg = Window.partitionBy('user_i', 'movieId').orderBy(\"userId\", \"movieId\")\n",
    "\n",
    "test_predictions = test_data.join(weight_data,test_data.userId == weight_data.user_i, \"inner\").drop('user_Id') \\\n",
    "    .withColumn(\"weighted_deviation\", col('weight') * col('calculations_j')[col('movieId')]) \\\n",
    "    .filter(col('weighted_deviation').isNotNull()) \\\n",
    "    .withColumn(\"numerator\", sum(col('weighted_deviation')).over(window) ) \\\n",
    "    .withColumn(\"denominator\", sum(abs(col('weight'))).over(window) ) \\\n",
    "    .withColumn(\"prediction\", col(\"numerator\") / col(\"denominator\") + col('calculations_i')['avg']) \\\n",
    "    .withColumn(\"row\", row_number().over(windowAgg)) \\\n",
    "    .filter(col(\"row\") == 1).drop(\"row\") \\\n",
    "    .select(\"userId\", \"movieId\", \"rating\", \"prediction\")"
   ],
   "metadata": {
    "collapsed": false
   }
  },
  {
   "cell_type": "code",
   "execution_count": 213,
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "                                                                                \r"
     ]
    },
    {
     "data": {
      "text/plain": "0.6041636794516598"
     },
     "execution_count": 213,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "test_predictions \\\n",
    "    .withColumn('error', (col('prediction') - col('rating')) * (col('prediction') - col('rating'))) \\\n",
    "    .select(mean(col('error'))) \\\n",
    "    .collect()[0][0]"
   ],
   "metadata": {
    "collapsed": false
   }
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "outputs": [],
   "source": [
    "# edges = weight_data.rdd.map(lambda row: (1, [(row[0], int(d['neighbor']), d['weight']) for d in row[1]]) if row[0] < 100 else (1, [])).flatMap(lambda row: row[1]).collect()"
   ],
   "metadata": {
    "collapsed": false
   }
  },
  {
   "cell_type": "code",
   "execution_count": 214,
   "outputs": [],
   "source": [
    "# G = nx.Graph()\n",
    "# G.add_weighted_edges_from(edges_test)\n",
    "# plt.figure(num=None, figsize=(30, 20), dpi=80)\n",
    "# plt.figure(num=None, figsize=(30, 20), dpi=80)\n",
    "#\n",
    "# pos = nx.spring_layout(G, k=0.5, iterations=50)\n",
    "# # nx.draw(G, node_size=350, font_size='large', with_labels=True, pos=pos)\n",
    "# nx.draw_networkx_nodes(G,pos, node_size=350, node_color='#6D9886')\n",
    "# nx.draw_networkx_edges(G,pos, edge_color='#393E46')\n",
    "# nx.draw_networkx_labels(G,pos, font_size='larger')\n",
    "# plt.axis('off')\n",
    "# plt.savefig('./images/users_relationship.png')\n",
    "# plt.show()"
   ],
   "metadata": {
    "collapsed": false
   }
  },
  {
   "cell_type": "code",
   "execution_count": 215,
   "outputs": [],
   "source": [
    "\n",
    "# G = nx.Graph()\n",
    "# G.add_weighted_edges_from(edges)\n",
    "# plt.figure(num=None, figsize=(30, 20), dpi=80)\n",
    "#\n",
    "# pos = nx.spring_layout(G, k=0.4, iterations=20)\n",
    "# # nx.draw(G, node_size=350, font_size='large', with_labels=True, pos=pos)\n",
    "# nx.draw_networkx_nodes(G,pos, node_size=70, node_color='#6D9886')\n",
    "# nx.draw_networkx_edges(G,pos, edge_color='#393E46')\n",
    "# plt.axis('off')\n",
    "# # plt.savefig('./images/users_relationship.png')\n",
    "# plt.show()"
   ],
   "metadata": {
    "collapsed": false
   }
  },
  {
   "cell_type": "code",
   "execution_count": 216,
   "outputs": [],
   "source": [
    "spark.stop()"
   ],
   "metadata": {
    "collapsed": false
   }
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 2
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython2",
   "version": "2.7.6"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 0
}