spotlight_implicit_sequence.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Spotlight Implicit Sequence\n",
    "\n",
    "A model for recommending items given a sequence of previous user item interactions.\n",
    "\n",
    "[Spotlight Documentation](https://maciejkula.github.io/spotlight/sequence/implicit.html)\n",
    "\n",
    "## Table of contents\n",
    "\n",
    "* [Sample files](#sample-files)\n",
    "* [Step 1 - Prepare training data](#prepare-training-data)\n",
    " * [Download movielens 100k dataset](#download-movielens)\n",
    " * [Import ratings data](#import-ratings-data)\n",
    " * [Create training data file](#create-training-data-file)\n",
    " * [Upload training data file](#upload-training-data)\n",
    "* [Step 2 - Create a model](#create-model)\n",
    " * [Run a SageMaker training job](#run-training-job)\n",
    " * [Create a SageMaker model](#create-sagemaker-model)\n",
    "* [Step 3 - Get recommendations (inference)](#get-recommendations)\n",
    " * [Import movie titles](#import-movie-titles)\n",
    " * [Example users](#example-users)\n",
    " * [Create sequences for prediction](#create-sequences)\n",
    " * [Create batch transform input file](#create-batch-input)\n",
    " * [Upload the batch transform input file to s3](#upload-batch-input)\n",
    " * [Run the Batch Transform Job](#run-transform)\n",
    " * [Download the batch results](#download-batch-results)\n",
    " * [Recommendations with scores](#recommendations)\n",
    " * [User history](#user-history)\n",
    "* [Step 4 - Optional Cleanup](#cleanup)\n",
    "\n",
    "## Sample files <a id=\"sample-files\"></a>\n",
    "\n",
    "These links are to example files on github.\n",
    "\n",
    "* [training input file](https://github.com/outpace/sagemaker-examples/blob/master/sequence_train_data/ml-100k-gt2.csv)\n",
    "* [batch transform input file](https://github.com/outpace/sagemaker-examples/blob/master/sequence_batch_input/recommendation.requests)\n",
    "* [batch transform output file](https://github.com/outpace/sagemaker-examples/blob/master/spotlight_sequence/recommendation.requests.out)\n",
    "\n",
    "## Step 1 - Prepare training data <a id=\"prepare-training-data\"></a>\n",
    "### Download movielens 100k dataset <a id=\"download-movielens\"></a>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "--2018-11-02 15:31:55--  http://files.grouplens.org/datasets/movielens/ml-100k.zip\n",
      "Resolving files.grouplens.org (files.grouplens.org)... 128.101.34.235\n",
      "Connecting to files.grouplens.org (files.grouplens.org)|128.101.34.235|:80... connected.\n",
      "HTTP request sent, awaiting response... 200 OK\n",
      "Length: 4924029 (4.7M) [application/zip]\n",
      "Saving to: ‘ml-100k.zip’\n",
      "\n",
      "ml-100k.zip         100%[===================>]   4.70M  18.5MB/s    in 0.3s    \n",
      "\n",
      "2018-11-02 15:31:55 (18.5 MB/s) - ‘ml-100k.zip’ saved [4924029/4924029]\n",
      "\n",
      "Archive:  ml-100k.zip\n",
      "   creating: ml-100k/\n",
      "  inflating: ml-100k/allbut.pl       \n",
      "  inflating: ml-100k/mku.sh          \n",
      "  inflating: ml-100k/README          \n",
      "  inflating: ml-100k/u.data          \n",
      "  inflating: ml-100k/u.genre         \n",
      "  inflating: ml-100k/u.info          \n",
      "  inflating: ml-100k/u.item          \n",
      "  inflating: ml-100k/u.occupation    \n",
      "  inflating: ml-100k/u.user          \n",
      "  inflating: ml-100k/u1.base         \n",
      "  inflating: ml-100k/u1.test         \n",
      "  inflating: ml-100k/u2.base         \n",
      "  inflating: ml-100k/u2.test         \n",
      "  inflating: ml-100k/u3.base         \n",
      "  inflating: ml-100k/u3.test         \n",
      "  inflating: ml-100k/u4.base         \n",
      "  inflating: ml-100k/u4.test         \n",
      "  inflating: ml-100k/u5.base         \n",
      "  inflating: ml-100k/u5.test         \n",
      "  inflating: ml-100k/ua.base         \n",
      "  inflating: ml-100k/ua.test         \n",
      "  inflating: ml-100k/ub.base         \n",
      "  inflating: ml-100k/ub.test         \n"
     ]
    }
   ],
   "source": [
    "!wget --no-clobber http://files.grouplens.org/datasets/movielens/ml-100k.zip\n",
    "!unzip -o ml-100k.zip"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Import ratings data <a id=\"import-ratings-data\"></a>\n",
    "\n",
    "Keep only ratings strictly higher than 2 to make this an implicit dataset."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "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>user_id</th>\n",
       "      <th>item_id</th>\n",
       "      <th>timestamp</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>196</td>\n",
       "      <td>242</td>\n",
       "      <td>881250949</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>186</td>\n",
       "      <td>302</td>\n",
       "      <td>891717742</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>298</td>\n",
       "      <td>474</td>\n",
       "      <td>884182806</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>253</td>\n",
       "      <td>465</td>\n",
       "      <td>891628467</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>305</td>\n",
       "      <td>451</td>\n",
       "      <td>886324817</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   user_id  item_id  timestamp\n",
       "0      196      242  881250949\n",
       "1      186      302  891717742\n",
       "5      298      474  884182806\n",
       "7      253      465  891628467\n",
       "8      305      451  886324817"
      ]
     },
     "execution_count": 2,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "import pandas as pd\n",
    "\n",
    "sequence_df = pd.read_csv('ml-100k/u.data', sep=\"\\t\", header=None, names=[\"user_id\", \"item_id\", \"rating\", \"timestamp\"])\n",
    "sequence_df = sequence_df[sequence_df.rating>2].drop([\"rating\"], axis=1)\n",
    "sequence_df.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Create training data file <a id=\"create-training-data-file\"></a>\n",
    "\n",
    "Create a csv file from the dataframe above. Do not include the index, but include headers `user_id`, `item_id`, and `timestamp`. Show the head of the file."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "user_id,item_id,timestamp\r\n",
      "196,242,881250949\r\n",
      "186,302,891717742\r\n",
      "298,474,884182806\r\n",
      "253,465,891628467\r\n",
      "305,451,886324817\r\n",
      "6,86,883603013\r\n",
      "286,1014,879781125\r\n",
      "200,222,876042340\r\n",
      "210,40,891035994\r\n"
     ]
    }
   ],
   "source": [
    "train_data_dir = 'sequence_train_data'\n",
    "train_data_file = '{}/ml-100k-gt2.csv'.format(train_data_dir)\n",
    "\n",
    "!mkdir -p {train_data_dir}\n",
    "sequence_df.to_csv(train_data_file, index=False)\n",
    "\n",
    "!head {train_data_file}"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Upload training data to s3 <a id=\"upload-training-data\"></a>\n",
    "\n",
    "Choose a bucket, optionally customize the prefix, and upload the csv created above."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'uploaded training data file to s3://sagemaker-validation-us-east-2/spotlight-implicit-sequence-test/training'"
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "import sagemaker as sage\n",
    "\n",
    "bucket = \"sagemaker-validation-us-east-2\"\n",
    "prefix = \"spotlight-implicit-sequence-test\"\n",
    "\n",
    "sess = sage.Session()\n",
    "\n",
    "s3_train = sess.upload_data(train_data_dir, bucket, \"{}/training\".format(prefix))\n",
    "\"uploaded training data file to {}\".format(s3_train)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Step 2 - Create a model <a id=\"create-model\"></a>\n",
    "\n",
    "### Run a SageMaker training job <a id=\"run-training-job\"></a>\n",
    "\n",
    "This code will start a training job, wait for it to be done, and report its status."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Training job current status: InProgress\n",
      "Training job ended with status: Completed\n",
      "CPU times: user 148 ms, sys: 13.6 ms, total: 161 ms\n",
      "Wall time: 4min\n"
     ]
    }
   ],
   "source": [
    "%%time\n",
    "\n",
    "import boto3\n",
    "import time\n",
    "from sagemaker import get_execution_role\n",
    "\n",
    "role = get_execution_role()\n",
    "ecr_image = \"435525115971.dkr.ecr.us-east-2.amazonaws.com/sagemaker/spotlight-sequence:6\"\n",
    "job_name_prefix = 'spotlight-implicit-sequence-test'\n",
    "timestamp = time.strftime('-%Y-%m-%d-%H-%M-%S', time.gmtime())\n",
    "job_name = job_name_prefix + timestamp\n",
    "create_training_params = \\\n",
    "{\n",
    "    \"AlgorithmSpecification\": {\n",
    "        \"TrainingImage\": ecr_image,\n",
    "        \"TrainingInputMode\": \"File\"\n",
    "    },\n",
    "    \"RoleArn\": role,\n",
    "    \"OutputDataConfig\": {\n",
    "        \"S3OutputPath\": 's3://{}/{}/output'.format(bucket, job_name_prefix)\n",
    "    },\n",
    "    \"ResourceConfig\": {\n",
    "        \"InstanceCount\": 1,\n",
    "        \"InstanceType\": \"ml.p3.2xlarge\",\n",
    "        \"VolumeSizeInGB\": 50\n",
    "    },\n",
    "    \"TrainingJobName\": job_name,\n",
    "    \"StoppingCondition\": {\n",
    "        \"MaxRuntimeInSeconds\": 360000\n",
    "    },\n",
    "    \"InputDataConfig\": [\n",
    "        {\n",
    "            \"ChannelName\": \"training\",\n",
    "            \"DataSource\": {\n",
    "                \"S3DataSource\": {\n",
    "                    \"S3DataType\": \"S3Prefix\",\n",
    "                    \"S3Uri\": s3_train,\n",
    "                    \"S3DataDistributionType\": \"FullyReplicated\"\n",
    "                }\n",
    "            },\n",
    "            \"ContentType\": \"application/csv\",\n",
    "            \"CompressionType\": \"None\"\n",
    "        }\n",
    "    ]\n",
    "}\n",
    "\n",
    "sagemaker = boto3.client(service_name='sagemaker')\n",
    "sagemaker.create_training_job(**create_training_params)\n",
    "status = sagemaker.describe_training_job(TrainingJobName=job_name)['TrainingJobStatus']\n",
    "print('Training job current status: {}'.format(status))\n",
    "\n",
    "try:\n",
    "    sagemaker.get_waiter('training_job_completed_or_stopped').wait(TrainingJobName=job_name)\n",
    "    job_info = sagemaker.describe_training_job(TrainingJobName=job_name)\n",
    "    status = job_info['TrainingJobStatus']\n",
    "    print(\"Training job ended with status: \" + status)\n",
    "except:\n",
    "    print('Training failed to start')\n",
    "    message = sagemaker.describe_training_job(TrainingJobName=job_name)['FailureReason']\n",
    "    print('Training failed with the following error: {}'.format(message))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Create a SageMaker model <a id=\"create-sagemaker-model\"></a>\n",
    "\n",
    "This will set up the model created during training within SageMaker to be used later for recommendations."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{'ModelArn': 'arn:aws:sagemaker:us-east-2:435525115971:model/spotlight-implicit-sequence-test-2018-11-02-15-35-58',\n",
       " 'ResponseMetadata': {'RequestId': '26e7d995-647c-4701-ae14-b9fefa5570d6',\n",
       "  'HTTPStatusCode': 200,\n",
       "  'HTTPHeaders': {'x-amzn-requestid': '26e7d995-647c-4701-ae14-b9fefa5570d6',\n",
       "   'content-type': 'application/x-amz-json-1.1',\n",
       "   'content-length': '114',\n",
       "   'date': 'Fri, 02 Nov 2018 15:35:58 GMT'},\n",
       "  'RetryAttempts': 0}}"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "timestamp = time.strftime('-%Y-%m-%d-%H-%M-%S', time.gmtime())\n",
    "model_name=\"spotlight-implicit-sequence-test\" + timestamp\n",
    "job_info = sagemaker.describe_training_job(TrainingJobName=job_name)\n",
    "model_data = job_info['ModelArtifacts']['S3ModelArtifacts']\n",
    "\n",
    "primary_container = {\n",
    "    'Image': ecr_image,\n",
    "    'ModelDataUrl': model_data,\n",
    "}\n",
    "\n",
    "create_model_response = sagemaker.create_model(\n",
    "    ModelName = model_name,\n",
    "    ExecutionRoleArn = role,\n",
    "    PrimaryContainer = primary_container)\n",
    "\n",
    "create_model_response"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Step 3 - Get recommendations (Inference) <a id=\"get-recommendations\"></a>\n",
    "\n",
    "### Import movie titles <a id=\"import-movie-titles\"></a>\n",
    "\n",
    "Get movie titles in `u.item` from the movielens files downloaded earlier and join with interactions data."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "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>user_id</th>\n",
       "      <th>item_id</th>\n",
       "      <th>timestamp</th>\n",
       "      <th>movie_title</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>196</td>\n",
       "      <td>242</td>\n",
       "      <td>881250949</td>\n",
       "      <td>Kolya (1996)</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>186</td>\n",
       "      <td>302</td>\n",
       "      <td>891717742</td>\n",
       "      <td>L.A. Confidential (1997)</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>298</td>\n",
       "      <td>474</td>\n",
       "      <td>884182806</td>\n",
       "      <td>Dr. Strangelove or: How I Learned to Stop Worr...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>253</td>\n",
       "      <td>465</td>\n",
       "      <td>891628467</td>\n",
       "      <td>Jungle Book, The (1994)</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>305</td>\n",
       "      <td>451</td>\n",
       "      <td>886324817</td>\n",
       "      <td>Grease (1978)</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   user_id  item_id  timestamp  \\\n",
       "0      196      242  881250949   \n",
       "1      186      302  891717742   \n",
       "5      298      474  884182806   \n",
       "7      253      465  891628467   \n",
       "8      305      451  886324817   \n",
       "\n",
       "                                         movie_title  \n",
       "0                                       Kolya (1996)  \n",
       "1                           L.A. Confidential (1997)  \n",
       "5  Dr. Strangelove or: How I Learned to Stop Worr...  \n",
       "7                            Jungle Book, The (1994)  \n",
       "8                                      Grease (1978)  "
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "titles_df = pd.read_csv('ml-100k/u.item', \n",
    "                        sep=\"|\", \n",
    "                        header=None, \n",
    "                        encoding = \"ISO-8859-1\"\n",
    "                       ).iloc[:,0:2].rename(index=str, columns={1:\"movie_title\"}).set_index([0])\n",
    "sequence_df = sequence_df.join(titles_df, on='item_id').rename(index=str, columns={\"user_id\":\"user_id\",1:\"movie_title\"})\n",
    "sequence_df.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Example users <a id=\"example-users\"></a>\n",
    "\n",
    "Find some example users to get their sequence of interactions in order to predict their next rating/watch."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "scrolled": false
   },
   "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>user_id</th>\n",
       "      <th>item_id</th>\n",
       "      <th>timestamp</th>\n",
       "      <th>movie_title</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>42723</th>\n",
       "      <td>685</td>\n",
       "      <td>269</td>\n",
       "      <td>879451401</td>\n",
       "      <td>Full Monty, The (1997)</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>43618</th>\n",
       "      <td>685</td>\n",
       "      <td>302</td>\n",
       "      <td>879451401</td>\n",
       "      <td>L.A. Confidential (1997)</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>53414</th>\n",
       "      <td>685</td>\n",
       "      <td>325</td>\n",
       "      <td>879451401</td>\n",
       "      <td>Crash (1996)</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>70388</th>\n",
       "      <td>685</td>\n",
       "      <td>324</td>\n",
       "      <td>879451401</td>\n",
       "      <td>Lost Highway (1997)</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>89437</th>\n",
       "      <td>685</td>\n",
       "      <td>882</td>\n",
       "      <td>879451401</td>\n",
       "      <td>Washington Square (1997)</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>98163</th>\n",
       "      <td>685</td>\n",
       "      <td>875</td>\n",
       "      <td>879451401</td>\n",
       "      <td>She's So Lovely (1997)</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>55280</th>\n",
       "      <td>302</td>\n",
       "      <td>358</td>\n",
       "      <td>879436981</td>\n",
       "      <td>Spawn (1997)</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>81824</th>\n",
       "      <td>302</td>\n",
       "      <td>271</td>\n",
       "      <td>879436911</td>\n",
       "      <td>Starship Troopers (1997)</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>58767</th>\n",
       "      <td>302</td>\n",
       "      <td>289</td>\n",
       "      <td>879436874</td>\n",
       "      <td>Evita (1996)</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4826</th>\n",
       "      <td>302</td>\n",
       "      <td>328</td>\n",
       "      <td>879436844</td>\n",
       "      <td>Conspiracy Theory (1997)</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>32327</th>\n",
       "      <td>302</td>\n",
       "      <td>301</td>\n",
       "      <td>879436820</td>\n",
       "      <td>In &amp; Out (1997)</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>84234</th>\n",
       "      <td>302</td>\n",
       "      <td>333</td>\n",
       "      <td>879436785</td>\n",
       "      <td>Game, The (1997)</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9848</th>\n",
       "      <td>302</td>\n",
       "      <td>307</td>\n",
       "      <td>879436739</td>\n",
       "      <td>Devil's Advocate, The (1997)</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>14758</th>\n",
       "      <td>302</td>\n",
       "      <td>258</td>\n",
       "      <td>879436739</td>\n",
       "      <td>Contact (1997)</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "       user_id  item_id  timestamp                   movie_title\n",
       "42723      685      269  879451401        Full Monty, The (1997)\n",
       "43618      685      302  879451401      L.A. Confidential (1997)\n",
       "53414      685      325  879451401                  Crash (1996)\n",
       "70388      685      324  879451401           Lost Highway (1997)\n",
       "89437      685      882  879451401      Washington Square (1997)\n",
       "98163      685      875  879451401        She's So Lovely (1997)\n",
       "55280      302      358  879436981                  Spawn (1997)\n",
       "81824      302      271  879436911      Starship Troopers (1997)\n",
       "58767      302      289  879436874                  Evita (1996)\n",
       "4826       302      328  879436844      Conspiracy Theory (1997)\n",
       "32327      302      301  879436820               In & Out (1997)\n",
       "84234      302      333  879436785              Game, The (1997)\n",
       "9848       302      307  879436739  Devil's Advocate, The (1997)\n",
       "14758      302      258  879436739                Contact (1997)"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "example_users = sequence_df[sequence_df.user_id.isin([685, 302])].sort_values(by=['timestamp'], ascending=False)\n",
    "example_users"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Create sequences for prediction <a id=\"create-sequences\"></a>\n",
    "\n",
    "Get each user's sequence of item_ids they interacted with."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([269, 302, 325, 324, 882, 875, 358, 271, 289, 328, 301, 333, 307,\n",
       "       258])"
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "user_685_sequence = example_users.sort_values(by=['timestamp'], ascending=False).item_id.values\n",
    "user_685_sequence"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([269, 302, 325, 324, 882, 875, 358, 271, 289, 328, 301, 333, 307,\n",
       "       258])"
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "user_302_sequence = example_users.sort_values(by=['timestamp'], ascending=False).item_id.values\n",
    "user_302_sequence"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Create batch transform input file <a id=\"create-batch-input\"></a>\n",
    "\n",
    "Each row is a json object containing three keys:\n",
    "\n",
    "* `sequence_id`: the id of the sequence - used to correlate results\n",
    "* `sequence`: the item ids a user has interacted with in order\n",
    "* `top_n`: the number of top scoring recommendations to return\n",
    "\n",
    "The head of the batch input file is shown."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "{\"sequence_id\": \"user_302\", \"sequence\": [\"269\", \"302\", \"325\", \"324\", \"882\", \"875\", \"358\", \"271\", \"289\", \"328\", \"301\", \"333\", \"307\", \"258\"], \"top_n\": \"5\"}\r\n",
      "{\"sequence_id\": \"user_685\", \"sequence\": [\"269\", \"302\", \"325\", \"324\", \"882\", \"875\", \"358\", \"271\", \"289\", \"328\", \"301\", \"333\", \"307\", \"258\"], \"top_n\": \"5\"}"
     ]
    }
   ],
   "source": [
    "import json\n",
    "\n",
    "batch_input_dir = 'sequence_batch_input'\n",
    "batch_input_file = batch_input_dir + '/recommendation.requests'\n",
    "\n",
    "!mkdir -p {batch_input_dir}\n",
    "\n",
    "with open(batch_input_file, 'w') as outfile:\n",
    "    json.dump({\"sequence_id\": \"user_302\", \"sequence\": [str(i) for i in user_302_sequence], \"top_n\": \"5\"}, outfile)\n",
    "    outfile.write(\"\\n\")\n",
    "    json.dump({\"sequence_id\": \"user_685\", \"sequence\": [str(i) for i in user_685_sequence], \"top_n\": \"5\"}, outfile)\n",
    "   \n",
    "!head {batch_input_file}"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Upload the batch transform input file to s3 <a id=\"upload-batch-input\"></a>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'uploaded training data file to s3://sagemaker-validation-us-east-2/spotlight-implicit-sequence-test/batch_input'"
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "batch_input = sess.upload_data(batch_input_dir, bucket, \"{}/batch_input\".format(prefix))\n",
    "\"uploaded training data file to {}\".format(batch_input)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Run the Batch Transform Job <a id=\"run-transform\"></a>\n",
    "\n",
    "This code will start a batch transform job, wait for it to be done, and report its status."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Created Transform job with name:  spotlight-implicit-sequence-test-2018-11-02-15-35-58\n",
      "Transform job ended with status: Completed\n",
      "CPU times: user 103 ms, sys: 3.93 ms, total: 107 ms\n",
      "Wall time: 4min 1s\n"
     ]
    }
   ],
   "source": [
    "%%time\n",
    "\n",
    "timestamp = time.strftime('-%Y-%m-%d-%H-%M-%S', time.gmtime())\n",
    "batch_job_name = \"spotlight-implicit-sequence-test\" + timestamp\n",
    "batch_output = 's3://{}/{}/output'.format(bucket, batch_job_name)\n",
    "request = \\\n",
    "{\n",
    "  \"TransformJobName\": batch_job_name,\n",
    "  \"ModelName\": model_name,\n",
    "  \"BatchStrategy\": \"SingleRecord\",\n",
    "  \"TransformInput\": {\n",
    "    \"DataSource\": {\n",
    "      \"S3DataSource\": {\n",
    "        \"S3DataType\": \"S3Prefix\",\n",
    "        \"S3Uri\": batch_input\n",
    "      }\n",
    "    },\n",
    "    \"ContentType\": \"application/json\",\n",
    "    \"CompressionType\": \"None\",\n",
    "    \"SplitType\": \"Line\"\n",
    "  },\n",
    "  \"TransformOutput\": {\n",
    "    \"S3OutputPath\": batch_output,\n",
    "    \"Accept\": \"text/csv\",\n",
    "    \"AssembleWith\": \"Line\"\n",
    "  },\n",
    "  \"TransformResources\": {\n",
    "    \"InstanceType\": \"ml.p3.2xlarge\",\n",
    "    \"InstanceCount\": 1\n",
    "  }\n",
    "}\n",
    "\n",
    "sagemaker.create_transform_job(**request)\n",
    "\n",
    "print(\"Created Transform job with name: \", batch_job_name)\n",
    "\n",
    "while(True):\n",
    "    job_info = sagemaker.describe_transform_job(TransformJobName=batch_job_name)\n",
    "    status = job_info['TransformJobStatus']\n",
    "    if status == 'Completed':\n",
    "        print(\"Transform job ended with status: \" + status)\n",
    "        break\n",
    "    if status == 'Failed':\n",
    "        message = job_info['FailureReason']\n",
    "        print('Transform failed with the following error: {}'.format(message))\n",
    "        raise Exception('Transform job failed') \n",
    "    time.sleep(30)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Download the batch results <a id=\"download-batch-results\"></a>\n",
    "\n",
    "Show the head of the file."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "download: s3://sagemaker-validation-us-east-2/spotlight-implicit-sequence-test-2018-11-02-15-35-58/output/recommendation.requests.out to spotlight_sequence/recommendation.requests.out\n",
      "user_302,313,33.03297424316406\n",
      "user_302,302,32.48430633544922\n",
      "user_302,268,31.736766815185547\n",
      "user_302,286,31.716033935546875\n",
      "user_302,300,31.430864334106445\n",
      "user_685,313,33.03297424316406\n",
      "user_685,302,32.48430633544922\n",
      "user_685,268,31.736766815185547\n",
      "user_685,286,31.716033935546875\n",
      "user_685,300,31.430864334106445\n"
     ]
    }
   ],
   "source": [
    "!aws s3 cp {batch_output + '/recommendation.requests.out'} spotlight_sequence/\n",
    "\n",
    "!head spotlight_sequence/recommendation.requests.out"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Recommendations with scores <a id=\"recommendations\"></a>\n",
    "\n",
    "Import the recommendations from the batch output file downloaded above and join with titles dataframe. These are the top 5 movie recommendations for users 685 and 302."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "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>sequence_id</th>\n",
       "      <th>item_id</th>\n",
       "      <th>score</th>\n",
       "      <th>movie_title</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>685</td>\n",
       "      <td>333</td>\n",
       "      <td>14.640368</td>\n",
       "      <td>Game, The (1997)</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>685</td>\n",
       "      <td>272</td>\n",
       "      <td>13.269605</td>\n",
       "      <td>Good Will Hunting (1997)</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>685</td>\n",
       "      <td>268</td>\n",
       "      <td>13.025604</td>\n",
       "      <td>Chasing Amy (1997)</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>685</td>\n",
       "      <td>347</td>\n",
       "      <td>12.637764</td>\n",
       "      <td>Wag the Dog (1997)</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>685</td>\n",
       "      <td>315</td>\n",
       "      <td>12.535725</td>\n",
       "      <td>Apt Pupil (1998)</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>302</td>\n",
       "      <td>748</td>\n",
       "      <td>12.058822</td>\n",
       "      <td>Saint, The (1997)</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>302</td>\n",
       "      <td>333</td>\n",
       "      <td>11.834901</td>\n",
       "      <td>Game, The (1997)</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>302</td>\n",
       "      <td>323</td>\n",
       "      <td>11.719286</td>\n",
       "      <td>Dante's Peak (1997)</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>302</td>\n",
       "      <td>258</td>\n",
       "      <td>10.959321</td>\n",
       "      <td>Contact (1997)</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>302</td>\n",
       "      <td>313</td>\n",
       "      <td>10.367278</td>\n",
       "      <td>Titanic (1997)</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   sequence_id  item_id      score               movie_title\n",
       "0          685      333  14.640368          Game, The (1997)\n",
       "1          685      272  13.269605  Good Will Hunting (1997)\n",
       "2          685      268  13.025604        Chasing Amy (1997)\n",
       "3          685      347  12.637764        Wag the Dog (1997)\n",
       "4          685      315  12.535725          Apt Pupil (1998)\n",
       "5          302      748  12.058822         Saint, The (1997)\n",
       "6          302      333  11.834901          Game, The (1997)\n",
       "7          302      323  11.719286       Dante's Peak (1997)\n",
       "8          302      258  10.959321            Contact (1997)\n",
       "9          302      313  10.367278            Titanic (1997)"
      ]
     },
     "execution_count": 15,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "recommendations_df = pd.read_csv('recommendation.requests.out', \n",
    "                                 header=None, \n",
    "                                 names=[\"sequence_id\", \"item_id\", \"score\"])\n",
    "recommendations_df = recommendations_df.join(titles_df, \n",
    "                                             on='item_id')\n",
    "recommendations_df"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### User history <a id=\"user-history\"></a>\n",
    "\n",
    "Show the example users interaction history again for convenience."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "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>user_id</th>\n",
       "      <th>item_id</th>\n",
       "      <th>timestamp</th>\n",
       "      <th>movie_title</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>42723</th>\n",
       "      <td>685</td>\n",
       "      <td>269</td>\n",
       "      <td>879451401</td>\n",
       "      <td>Full Monty, The (1997)</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>43618</th>\n",
       "      <td>685</td>\n",
       "      <td>302</td>\n",
       "      <td>879451401</td>\n",
       "      <td>L.A. Confidential (1997)</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>53414</th>\n",
       "      <td>685</td>\n",
       "      <td>325</td>\n",
       "      <td>879451401</td>\n",
       "      <td>Crash (1996)</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>70388</th>\n",
       "      <td>685</td>\n",
       "      <td>324</td>\n",
       "      <td>879451401</td>\n",
       "      <td>Lost Highway (1997)</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>89437</th>\n",
       "      <td>685</td>\n",
       "      <td>882</td>\n",
       "      <td>879451401</td>\n",
       "      <td>Washington Square (1997)</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>98163</th>\n",
       "      <td>685</td>\n",
       "      <td>875</td>\n",
       "      <td>879451401</td>\n",
       "      <td>She's So Lovely (1997)</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>55280</th>\n",
       "      <td>302</td>\n",
       "      <td>358</td>\n",
       "      <td>879436981</td>\n",
       "      <td>Spawn (1997)</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>81824</th>\n",
       "      <td>302</td>\n",
       "      <td>271</td>\n",
       "      <td>879436911</td>\n",
       "      <td>Starship Troopers (1997)</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>58767</th>\n",
       "      <td>302</td>\n",
       "      <td>289</td>\n",
       "      <td>879436874</td>\n",
       "      <td>Evita (1996)</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4826</th>\n",
       "      <td>302</td>\n",
       "      <td>328</td>\n",
       "      <td>879436844</td>\n",
       "      <td>Conspiracy Theory (1997)</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>32327</th>\n",
       "      <td>302</td>\n",
       "      <td>301</td>\n",
       "      <td>879436820</td>\n",
       "      <td>In &amp; Out (1997)</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>84234</th>\n",
       "      <td>302</td>\n",
       "      <td>333</td>\n",
       "      <td>879436785</td>\n",
       "      <td>Game, The (1997)</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9848</th>\n",
       "      <td>302</td>\n",
       "      <td>307</td>\n",
       "      <td>879436739</td>\n",
       "      <td>Devil's Advocate, The (1997)</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>14758</th>\n",
       "      <td>302</td>\n",
       "      <td>258</td>\n",
       "      <td>879436739</td>\n",
       "      <td>Contact (1997)</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "       user_id  item_id  timestamp                   movie_title\n",
       "42723      685      269  879451401        Full Monty, The (1997)\n",
       "43618      685      302  879451401      L.A. Confidential (1997)\n",
       "53414      685      325  879451401                  Crash (1996)\n",
       "70388      685      324  879451401           Lost Highway (1997)\n",
       "89437      685      882  879451401      Washington Square (1997)\n",
       "98163      685      875  879451401        She's So Lovely (1997)\n",
       "55280      302      358  879436981                  Spawn (1997)\n",
       "81824      302      271  879436911      Starship Troopers (1997)\n",
       "58767      302      289  879436874                  Evita (1996)\n",
       "4826       302      328  879436844      Conspiracy Theory (1997)\n",
       "32327      302      301  879436820               In & Out (1997)\n",
       "84234      302      333  879436785              Game, The (1997)\n",
       "9848       302      307  879436739  Devil's Advocate, The (1997)\n",
       "14758      302      258  879436739                Contact (1997)"
      ]
     },
     "execution_count": 16,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "example_users"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Step 4 - Optional Clean up <a id=\"cleanup\"></a>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [],
   "source": [
    "# optionally uncomment and run the code to clean everything up\n",
    "\n",
    "#!rm ml-100k.zip 2> /dev/null\n",
    "#!rm -fr spotlight_sequence/ 2> /dev/null\n",
    "#!rm -fr ml-100k/ 2> /dev/null\n",
    "#!rm -fr {train_data_dir} 2> /dev/null\n",
    "#!rm -fr {batch_input_dir} 2> /dev/null\n",
    "#sagemaker.delete_model(ModelName= model_name)"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "conda_pytorch_p36",
   "language": "python",
   "name": "conda_pytorch_p36"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.6.5"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}