whizlabs-course/lab/clustering-lab/clustering-kmeans.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import os\n",
    "import boto3\n",
    "import io\n",
    "import sagemaker\n",
    "\n",
    "%matplotlib inline \n",
    "\n",
    "import pandas as pd\n",
    "import numpy as np\n",
    "import mxnet as mx\n",
    "import matplotlib.pyplot as plt\n",
    "import matplotlib\n",
    "import seaborn as sns\n",
    "matplotlib.style.use('ggplot')\n",
    "import pickle, gzip, urllib, json\n",
    "import csv\n",
    "from sagemaker import get_execution_role\n",
    "role = get_execution_role()\n",
    "s3 = boto3.resource('s3')\n",
    "bucket_name = 'machine-learning-exam' # place the USCensus1990.data.csv file in a bucket in your account\n",
    "object_key = 'Absenteeism_at_work.csv'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Load our absenteeism data\n",
    "s3_client = boto3.client('s3')\n",
    "response = s3_client.get_object(Bucket=bucket_name, Key=object_key)\n",
    "response_body = response[\"Body\"].read()\n",
    "absence = pd.read_csv(io.BytesIO(response_body), header=0, delimiter=\",\", low_memory=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "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>ID</th>\n",
       "      <th>Reason for absence</th>\n",
       "      <th>Month of absence</th>\n",
       "      <th>Day of the week</th>\n",
       "      <th>Seasons</th>\n",
       "      <th>Transportation expense</th>\n",
       "      <th>Distance from Residence to Work</th>\n",
       "      <th>Service time</th>\n",
       "      <th>Age</th>\n",
       "      <th>Work load Average/day</th>\n",
       "      <th>...</th>\n",
       "      <th>Disciplinary failure</th>\n",
       "      <th>Education</th>\n",
       "      <th>Son</th>\n",
       "      <th>Social drinker</th>\n",
       "      <th>Social smoker</th>\n",
       "      <th>Pet</th>\n",
       "      <th>Weight</th>\n",
       "      <th>Height</th>\n",
       "      <th>Body mass index</th>\n",
       "      <th>Absenteeism time in hours</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>11</td>\n",
       "      <td>26</td>\n",
       "      <td>7</td>\n",
       "      <td>3</td>\n",
       "      <td>1</td>\n",
       "      <td>289</td>\n",
       "      <td>36</td>\n",
       "      <td>13</td>\n",
       "      <td>33</td>\n",
       "      <td>239.554</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>2</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>90</td>\n",
       "      <td>172</td>\n",
       "      <td>30</td>\n",
       "      <td>4</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>36</td>\n",
       "      <td>0</td>\n",
       "      <td>7</td>\n",
       "      <td>3</td>\n",
       "      <td>1</td>\n",
       "      <td>118</td>\n",
       "      <td>13</td>\n",
       "      <td>18</td>\n",
       "      <td>50</td>\n",
       "      <td>239.554</td>\n",
       "      <td>...</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>98</td>\n",
       "      <td>178</td>\n",
       "      <td>31</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>3</td>\n",
       "      <td>23</td>\n",
       "      <td>7</td>\n",
       "      <td>4</td>\n",
       "      <td>1</td>\n",
       "      <td>179</td>\n",
       "      <td>51</td>\n",
       "      <td>18</td>\n",
       "      <td>38</td>\n",
       "      <td>239.554</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>89</td>\n",
       "      <td>170</td>\n",
       "      <td>31</td>\n",
       "      <td>2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>7</td>\n",
       "      <td>7</td>\n",
       "      <td>7</td>\n",
       "      <td>5</td>\n",
       "      <td>1</td>\n",
       "      <td>279</td>\n",
       "      <td>5</td>\n",
       "      <td>14</td>\n",
       "      <td>39</td>\n",
       "      <td>239.554</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>2</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>68</td>\n",
       "      <td>168</td>\n",
       "      <td>24</td>\n",
       "      <td>4</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>11</td>\n",
       "      <td>23</td>\n",
       "      <td>7</td>\n",
       "      <td>5</td>\n",
       "      <td>1</td>\n",
       "      <td>289</td>\n",
       "      <td>36</td>\n",
       "      <td>13</td>\n",
       "      <td>33</td>\n",
       "      <td>239.554</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>2</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>90</td>\n",
       "      <td>172</td>\n",
       "      <td>30</td>\n",
       "      <td>2</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>5 rows × 21 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "   ID  Reason for absence  Month of absence  Day of the week  Seasons  \\\n",
       "0  11                  26                 7                3        1   \n",
       "1  36                   0                 7                3        1   \n",
       "2   3                  23                 7                4        1   \n",
       "3   7                   7                 7                5        1   \n",
       "4  11                  23                 7                5        1   \n",
       "\n",
       "   Transportation expense  Distance from Residence to Work  Service time  Age  \\\n",
       "0                     289                               36            13   33   \n",
       "1                     118                               13            18   50   \n",
       "2                     179                               51            18   38   \n",
       "3                     279                                5            14   39   \n",
       "4                     289                               36            13   33   \n",
       "\n",
       "   Work load Average/day   ...  Disciplinary failure  Education  Son  \\\n",
       "0                 239.554  ...                     0          1    2   \n",
       "1                 239.554  ...                     1          1    1   \n",
       "2                 239.554  ...                     0          1    0   \n",
       "3                 239.554  ...                     0          1    2   \n",
       "4                 239.554  ...                     0          1    2   \n",
       "\n",
       "   Social drinker  Social smoker  Pet  Weight  Height  Body mass index  \\\n",
       "0               1              0    1      90     172               30   \n",
       "1               1              0    0      98     178               31   \n",
       "2               1              0    0      89     170               31   \n",
       "3               1              1    0      68     168               24   \n",
       "4               1              0    1      90     172               30   \n",
       "\n",
       "   Absenteeism time in hours  \n",
       "0                          4  \n",
       "1                          0  \n",
       "2                          2  \n",
       "3                          4  \n",
       "4                          2  \n",
       "\n",
       "[5 rows x 21 columns]"
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "absence.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Drop the ID column\n",
    "absence = absence.drop(['ID'], axis=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(740, 20)"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "absence.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(740, 20)"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Drop any missing value observations\n",
    "absence.dropna(inplace=True)\n",
    "absence.shape"
   ]
  },
  {
   "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>Reason for absence</th>\n",
       "      <th>Month of absence</th>\n",
       "      <th>Day of the week</th>\n",
       "      <th>Seasons</th>\n",
       "      <th>Transportation expense</th>\n",
       "      <th>Distance from Residence to Work</th>\n",
       "      <th>Service time</th>\n",
       "      <th>Age</th>\n",
       "      <th>Work load Average/day</th>\n",
       "      <th>Hit target</th>\n",
       "      <th>Disciplinary failure</th>\n",
       "      <th>Education</th>\n",
       "      <th>Son</th>\n",
       "      <th>Social drinker</th>\n",
       "      <th>Social smoker</th>\n",
       "      <th>Pet</th>\n",
       "      <th>Weight</th>\n",
       "      <th>Height</th>\n",
       "      <th>Body mass index</th>\n",
       "      <th>Absenteeism time in hours</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Reason for absence</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>26</th>\n",
       "      <td>26</td>\n",
       "      <td>7</td>\n",
       "      <td>3</td>\n",
       "      <td>1</td>\n",
       "      <td>289</td>\n",
       "      <td>36</td>\n",
       "      <td>13</td>\n",
       "      <td>33</td>\n",
       "      <td>239.554</td>\n",
       "      <td>97</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>2</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>90</td>\n",
       "      <td>172</td>\n",
       "      <td>30</td>\n",
       "      <td>4</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0</td>\n",
       "      <td>7</td>\n",
       "      <td>3</td>\n",
       "      <td>1</td>\n",
       "      <td>118</td>\n",
       "      <td>13</td>\n",
       "      <td>18</td>\n",
       "      <td>50</td>\n",
       "      <td>239.554</td>\n",
       "      <td>97</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>98</td>\n",
       "      <td>178</td>\n",
       "      <td>31</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>23</th>\n",
       "      <td>23</td>\n",
       "      <td>7</td>\n",
       "      <td>4</td>\n",
       "      <td>1</td>\n",
       "      <td>179</td>\n",
       "      <td>51</td>\n",
       "      <td>18</td>\n",
       "      <td>38</td>\n",
       "      <td>239.554</td>\n",
       "      <td>97</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>89</td>\n",
       "      <td>170</td>\n",
       "      <td>31</td>\n",
       "      <td>2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>7</td>\n",
       "      <td>7</td>\n",
       "      <td>5</td>\n",
       "      <td>1</td>\n",
       "      <td>279</td>\n",
       "      <td>5</td>\n",
       "      <td>14</td>\n",
       "      <td>39</td>\n",
       "      <td>239.554</td>\n",
       "      <td>97</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>2</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>68</td>\n",
       "      <td>168</td>\n",
       "      <td>24</td>\n",
       "      <td>4</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>23</th>\n",
       "      <td>23</td>\n",
       "      <td>7</td>\n",
       "      <td>5</td>\n",
       "      <td>1</td>\n",
       "      <td>289</td>\n",
       "      <td>36</td>\n",
       "      <td>13</td>\n",
       "      <td>33</td>\n",
       "      <td>239.554</td>\n",
       "      <td>97</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>2</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>90</td>\n",
       "      <td>172</td>\n",
       "      <td>30</td>\n",
       "      <td>2</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                    Reason for absence  Month of absence  Day of the week  \\\n",
       "Reason for absence                                                          \n",
       "26                                  26                 7                3   \n",
       "0                                    0                 7                3   \n",
       "23                                  23                 7                4   \n",
       "7                                    7                 7                5   \n",
       "23                                  23                 7                5   \n",
       "\n",
       "                    Seasons  Transportation expense  \\\n",
       "Reason for absence                                    \n",
       "26                        1                     289   \n",
       "0                         1                     118   \n",
       "23                        1                     179   \n",
       "7                         1                     279   \n",
       "23                        1                     289   \n",
       "\n",
       "                    Distance from Residence to Work  Service time  Age  \\\n",
       "Reason for absence                                                       \n",
       "26                                               36            13   33   \n",
       "0                                                13            18   50   \n",
       "23                                               51            18   38   \n",
       "7                                                 5            14   39   \n",
       "23                                               36            13   33   \n",
       "\n",
       "                    Work load Average/day   Hit target  Disciplinary failure  \\\n",
       "Reason for absence                                                             \n",
       "26                                 239.554          97                     0   \n",
       "0                                  239.554          97                     1   \n",
       "23                                 239.554          97                     0   \n",
       "7                                  239.554          97                     0   \n",
       "23                                 239.554          97                     0   \n",
       "\n",
       "                    Education  Son  Social drinker  Social smoker  Pet  \\\n",
       "Reason for absence                                                       \n",
       "26                          1    2               1              0    1   \n",
       "0                           1    1               1              0    0   \n",
       "23                          1    0               1              0    0   \n",
       "7                           1    2               1              1    0   \n",
       "23                          1    2               1              0    1   \n",
       "\n",
       "                    Weight  Height  Body mass index  Absenteeism time in hours  \n",
       "Reason for absence                                                              \n",
       "26                      90     172               30                          4  \n",
       "0                       98     178               31                          0  \n",
       "23                      89     170               31                          2  \n",
       "7                       68     168               24                          4  \n",
       "23                      90     172               30                          2  "
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Set the 'Reason for absence' as the index \n",
    "# and the rest of the numerical features become the attributes of each unique absence\n",
    "absence.index=absence['Reason for absence']\n",
    "absence.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(740, 20)"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "absence.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/ec2-user/anaconda3/envs/python3/lib/python3.6/site-packages/scipy/stats/stats.py:1713: FutureWarning: Using a non-tuple sequence for multidimensional indexing is deprecated; use `arr[tuple(seq)]` instead of `arr[seq]`. In the future this will be interpreted as an array index, `arr[np.array(seq)]`, which will result either in an error or a different result.\n",
      "  return np.add.reduce(sorted[indexer] * weights, axis=axis) / sumval\n",
      "/home/ec2-user/anaconda3/envs/python3/lib/python3.6/site-packages/matplotlib/axes/_axes.py:6521: MatplotlibDeprecationWarning: \n",
      "The 'normed' kwarg was deprecated in Matplotlib 2.1 and will be removed in 3.1. Use 'density' instead.\n",
      "  alternative=\"'density'\", removal=\"3.1\")\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x216 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/ec2-user/anaconda3/envs/python3/lib/python3.6/site-packages/matplotlib/axes/_axes.py:6521: MatplotlibDeprecationWarning: \n",
      "The 'normed' kwarg was deprecated in Matplotlib 2.1 and will be removed in 3.1. Use 'density' instead.\n",
      "  alternative=\"'density'\", removal=\"3.1\")\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x216 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/ec2-user/anaconda3/envs/python3/lib/python3.6/site-packages/matplotlib/axes/_axes.py:6521: MatplotlibDeprecationWarning: \n",
      "The 'normed' kwarg was deprecated in Matplotlib 2.1 and will be removed in 3.1. Use 'density' instead.\n",
      "  alternative=\"'density'\", removal=\"3.1\")\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x216 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Visualize the data for some of our columns to see what the distribution looks like\n",
    "import seaborn as sns\n",
    "\n",
    "for a in ['Reason for absence', 'Transportation expense', 'Distance from Residence to Work']:\n",
    "    ax=plt.subplots(figsize=(6,3))\n",
    "    ax=sns.distplot(absence[a])\n",
    "    title=\"Histogram of \" + a\n",
    "    ax.set_title(title, fontsize=12)\n",
    "    plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/ec2-user/anaconda3/envs/python3/lib/python3.6/site-packages/sklearn/preprocessing/data.py:334: DataConversionWarning: Data with input dtype int64, float64 were all converted to float64 by MinMaxScaler.\n",
      "  return self.partial_fit(X, y)\n"
     ]
    }
   ],
   "source": [
    "# Standardize the scaling of the columns in order to use any distance based analytical methods \n",
    "# so that we can compare the relative distances between different feature columns\n",
    "# Use minmaxscaler to transform the columns so that they also fall between 0 and 1\n",
    "from sklearn.preprocessing import MinMaxScaler\n",
    "scaler=MinMaxScaler()\n",
    "absence_scaled=pd.DataFrame(scaler.fit_transform(absence))\n",
    "absence_scaled.columns=absence.columns\n",
    "absence_scaled.index=absence.index"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "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>Reason for absence</th>\n",
       "      <th>Month of absence</th>\n",
       "      <th>Day of the week</th>\n",
       "      <th>Seasons</th>\n",
       "      <th>Transportation expense</th>\n",
       "      <th>Distance from Residence to Work</th>\n",
       "      <th>Service time</th>\n",
       "      <th>Age</th>\n",
       "      <th>Work load Average/day</th>\n",
       "      <th>Hit target</th>\n",
       "      <th>Disciplinary failure</th>\n",
       "      <th>Education</th>\n",
       "      <th>Son</th>\n",
       "      <th>Social drinker</th>\n",
       "      <th>Social smoker</th>\n",
       "      <th>Pet</th>\n",
       "      <th>Weight</th>\n",
       "      <th>Height</th>\n",
       "      <th>Body mass index</th>\n",
       "      <th>Absenteeism time in hours</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>count</th>\n",
       "      <td>740.000000</td>\n",
       "      <td>740.000000</td>\n",
       "      <td>740.000000</td>\n",
       "      <td>740.000000</td>\n",
       "      <td>740.000000</td>\n",
       "      <td>740.000000</td>\n",
       "      <td>740.000000</td>\n",
       "      <td>740.000000</td>\n",
       "      <td>740.000000</td>\n",
       "      <td>740.000000</td>\n",
       "      <td>740.000000</td>\n",
       "      <td>740.000000</td>\n",
       "      <td>740.000000</td>\n",
       "      <td>740.000000</td>\n",
       "      <td>740.000000</td>\n",
       "      <td>740.000000</td>\n",
       "      <td>740.000000</td>\n",
       "      <td>740.000000</td>\n",
       "      <td>740.000000</td>\n",
       "      <td>740.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>mean</th>\n",
       "      <td>0.686293</td>\n",
       "      <td>0.527027</td>\n",
       "      <td>0.478716</td>\n",
       "      <td>0.514865</td>\n",
       "      <td>0.382703</td>\n",
       "      <td>0.524066</td>\n",
       "      <td>0.412645</td>\n",
       "      <td>0.304839</td>\n",
       "      <td>0.379108</td>\n",
       "      <td>0.715149</td>\n",
       "      <td>0.054054</td>\n",
       "      <td>0.097297</td>\n",
       "      <td>0.254730</td>\n",
       "      <td>0.567568</td>\n",
       "      <td>0.072973</td>\n",
       "      <td>0.093243</td>\n",
       "      <td>0.442983</td>\n",
       "      <td>0.276208</td>\n",
       "      <td>0.404054</td>\n",
       "      <td>0.057703</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>std</th>\n",
       "      <td>0.301193</td>\n",
       "      <td>0.286357</td>\n",
       "      <td>0.355419</td>\n",
       "      <td>0.370610</td>\n",
       "      <td>0.247971</td>\n",
       "      <td>0.315676</td>\n",
       "      <td>0.156603</td>\n",
       "      <td>0.208993</td>\n",
       "      <td>0.225813</td>\n",
       "      <td>0.198911</td>\n",
       "      <td>0.226277</td>\n",
       "      <td>0.224413</td>\n",
       "      <td>0.274622</td>\n",
       "      <td>0.495749</td>\n",
       "      <td>0.260268</td>\n",
       "      <td>0.164782</td>\n",
       "      <td>0.247754</td>\n",
       "      <td>0.182879</td>\n",
       "      <td>0.225550</td>\n",
       "      <td>0.111092</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>min</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>25%</th>\n",
       "      <td>0.464286</td>\n",
       "      <td>0.250000</td>\n",
       "      <td>0.250000</td>\n",
       "      <td>0.333333</td>\n",
       "      <td>0.225926</td>\n",
       "      <td>0.234043</td>\n",
       "      <td>0.285714</td>\n",
       "      <td>0.129032</td>\n",
       "      <td>0.222412</td>\n",
       "      <td>0.631579</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.250000</td>\n",
       "      <td>0.181818</td>\n",
       "      <td>0.263158</td>\n",
       "      <td>0.016667</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>50%</th>\n",
       "      <td>0.821429</td>\n",
       "      <td>0.500000</td>\n",
       "      <td>0.500000</td>\n",
       "      <td>0.666667</td>\n",
       "      <td>0.396296</td>\n",
       "      <td>0.446809</td>\n",
       "      <td>0.428571</td>\n",
       "      <td>0.322581</td>\n",
       "      <td>0.337244</td>\n",
       "      <td>0.736842</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.250000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.519231</td>\n",
       "      <td>0.212121</td>\n",
       "      <td>0.315789</td>\n",
       "      <td>0.025000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>75%</th>\n",
       "      <td>0.928571</td>\n",
       "      <td>0.750000</td>\n",
       "      <td>0.750000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.525926</td>\n",
       "      <td>0.957447</td>\n",
       "      <td>0.535714</td>\n",
       "      <td>0.419355</td>\n",
       "      <td>0.510502</td>\n",
       "      <td>0.842105</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.500000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.125000</td>\n",
       "      <td>0.634615</td>\n",
       "      <td>0.272727</td>\n",
       "      <td>0.631579</td>\n",
       "      <td>0.066667</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>max</th>\n",
       "      <td>1.000000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>1.000000</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "       Reason for absence  Month of absence  Day of the week     Seasons  \\\n",
       "count          740.000000        740.000000       740.000000  740.000000   \n",
       "mean             0.686293          0.527027         0.478716    0.514865   \n",
       "std              0.301193          0.286357         0.355419    0.370610   \n",
       "min              0.000000          0.000000         0.000000    0.000000   \n",
       "25%              0.464286          0.250000         0.250000    0.333333   \n",
       "50%              0.821429          0.500000         0.500000    0.666667   \n",
       "75%              0.928571          0.750000         0.750000    1.000000   \n",
       "max              1.000000          1.000000         1.000000    1.000000   \n",
       "\n",
       "       Transportation expense  Distance from Residence to Work  Service time  \\\n",
       "count              740.000000                       740.000000    740.000000   \n",
       "mean                 0.382703                         0.524066      0.412645   \n",
       "std                  0.247971                         0.315676      0.156603   \n",
       "min                  0.000000                         0.000000      0.000000   \n",
       "25%                  0.225926                         0.234043      0.285714   \n",
       "50%                  0.396296                         0.446809      0.428571   \n",
       "75%                  0.525926                         0.957447      0.535714   \n",
       "max                  1.000000                         1.000000      1.000000   \n",
       "\n",
       "              Age  Work load Average/day   Hit target  Disciplinary failure  \\\n",
       "count  740.000000              740.000000  740.000000            740.000000   \n",
       "mean     0.304839                0.379108    0.715149              0.054054   \n",
       "std      0.208993                0.225813    0.198911              0.226277   \n",
       "min      0.000000                0.000000    0.000000              0.000000   \n",
       "25%      0.129032                0.222412    0.631579              0.000000   \n",
       "50%      0.322581                0.337244    0.736842              0.000000   \n",
       "75%      0.419355                0.510502    0.842105              0.000000   \n",
       "max      1.000000                1.000000    1.000000              1.000000   \n",
       "\n",
       "        Education         Son  Social drinker  Social smoker         Pet  \\\n",
       "count  740.000000  740.000000      740.000000     740.000000  740.000000   \n",
       "mean     0.097297    0.254730        0.567568       0.072973    0.093243   \n",
       "std      0.224413    0.274622        0.495749       0.260268    0.164782   \n",
       "min      0.000000    0.000000        0.000000       0.000000    0.000000   \n",
       "25%      0.000000    0.000000        0.000000       0.000000    0.000000   \n",
       "50%      0.000000    0.250000        1.000000       0.000000    0.000000   \n",
       "75%      0.000000    0.500000        1.000000       0.000000    0.125000   \n",
       "max      1.000000    1.000000        1.000000       1.000000    1.000000   \n",
       "\n",
       "           Weight      Height  Body mass index  Absenteeism time in hours  \n",
       "count  740.000000  740.000000       740.000000                 740.000000  \n",
       "mean     0.442983    0.276208         0.404054                   0.057703  \n",
       "std      0.247754    0.182879         0.225550                   0.111092  \n",
       "min      0.000000    0.000000         0.000000                   0.000000  \n",
       "25%      0.250000    0.181818         0.263158                   0.016667  \n",
       "50%      0.519231    0.212121         0.315789                   0.025000  \n",
       "75%      0.634615    0.272727         0.631579                   0.066667  \n",
       "max      1.000000    1.000000         1.000000                   1.000000  "
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# See that all of our columns are now between 0 and 1\n",
    "absence_scaled.describe()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Use principal component analysis (PCA) to reduce the dimensionality of our data \n",
    "bucket='machine-learning-exam'\n",
    "from sagemaker import PCA\n",
    "num_components=19\n",
    "\n",
    "pca_SM = PCA(role=role,\n",
    "             train_instance_count=1,\n",
    "             train_instance_type='ml.c4.xlarge',\n",
    "             output_path='s3://'+ bucket +'/absence/',\n",
    "             num_components=num_components)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Extract the numpy array from the DataFrame and explicitly casting to float32\n",
    "train_data = absence_scaled.values.astype('float32')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "2020-02-18 02:03:36 Starting - Starting the training job...\n",
      "2020-02-18 02:03:37 Starting - Launching requested ML instances......\n",
      "2020-02-18 02:04:41 Starting - Preparing the instances for training......\n",
      "2020-02-18 02:06:02 Downloading - Downloading input data\n",
      "2020-02-18 02:06:02 Training - Downloading the training image..\u001b[34mDocker entrypoint called with argument(s): train\u001b[0m\n",
      "\u001b[34m[02/18/2020 02:06:18 INFO 140378775521088] Reading default configuration from /opt/amazon/lib/python2.7/site-packages/algorithm/resources/default-conf.json: {u'_num_gpus': u'auto', u'_log_level': u'info', u'subtract_mean': u'true', u'force_dense': u'true', u'epochs': 1, u'algorithm_mode': u'regular', u'extra_components': u'-1', u'_kvstore': u'dist_sync', u'_num_kv_servers': u'auto'}\u001b[0m\n",
      "\u001b[34m[02/18/2020 02:06:18 INFO 140378775521088] Reading provided configuration from /opt/ml/input/config/hyperparameters.json: {u'feature_dim': u'20', u'mini_batch_size': u'500', u'num_components': u'19'}\u001b[0m\n",
      "\u001b[34m[02/18/2020 02:06:18 INFO 140378775521088] Final configuration: {u'num_components': u'19', u'_num_gpus': u'auto', u'_log_level': u'info', u'subtract_mean': u'true', u'force_dense': u'true', u'epochs': 1, u'algorithm_mode': u'regular', u'feature_dim': u'20', u'extra_components': u'-1', u'_kvstore': u'dist_sync', u'_num_kv_servers': u'auto', u'mini_batch_size': u'500'}\u001b[0m\n",
      "\u001b[34m[02/18/2020 02:06:18 WARNING 140378775521088] Loggers have already been setup.\u001b[0m\n",
      "\u001b[34m[02/18/2020 02:06:18 INFO 140378775521088] Launching parameter server for role scheduler\u001b[0m\n",
      "\u001b[34m[02/18/2020 02:06:18 INFO 140378775521088] {'ECS_CONTAINER_METADATA_URI': 'http://169.254.170.2/v3/47ec9ef1-5bf3-4241-847a-f87ce6e331de', 'PROTOCOL_BUFFERS_PYTHON_IMPLEMENTATION_VERSION': '2', 'PATH': '/opt/amazon/bin:/usr/local/nvidia/bin:/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin:/opt/amazon/bin:/opt/amazon/bin', 'SAGEMAKER_HTTP_PORT': '8080', 'HOME': '/root', 'PYTHONUNBUFFERED': 'TRUE', 'CANONICAL_ENVROOT': '/opt/amazon', 'LD_LIBRARY_PATH': '/opt/amazon/lib/python2.7/site-packages/cv2/../../../../lib:/usr/local/nvidia/lib64:/opt/amazon/lib', 'LANG': 'en_US.utf8', 'DMLC_INTERFACE': 'eth0', 'SHLVL': '1', 'AWS_REGION': 'us-east-1', 'NVIDIA_VISIBLE_DEVICES': 'void', 'TRAINING_JOB_NAME': 'pca-2020-02-18-02-03-36-037', 'PROTOCOL_BUFFERS_PYTHON_IMPLEMENTATION': 'cpp', 'ENVROOT': '/opt/amazon', 'SAGEMAKER_DATA_PATH': '/opt/ml', 'NVIDIA_DRIVER_CAPABILITIES': 'compute,utility', 'NVIDIA_REQUIRE_CUDA': 'cuda>=9.0', 'OMP_NUM_THREADS': '2', 'HOSTNAME': 'ip-10-2-95-208.ec2.internal', 'AWS_CONTAINER_CREDENTIALS_RELATIVE_URI': '/v2/credentials/642304cc-1ed4-4342-b0f6-b9c4f36bbe0a', 'PWD': '/', 'TRAINING_JOB_ARN': 'arn:aws:sagemaker:us-east-1:001178231653:training-job/pca-2020-02-18-02-03-36-037', 'AWS_EXECUTION_ENV': 'AWS_ECS_EC2'}\u001b[0m\n",
      "\u001b[34m[02/18/2020 02:06:18 INFO 140378775521088] envs={'ECS_CONTAINER_METADATA_URI': 'http://169.254.170.2/v3/47ec9ef1-5bf3-4241-847a-f87ce6e331de', 'PROTOCOL_BUFFERS_PYTHON_IMPLEMENTATION_VERSION': '2', 'DMLC_NUM_WORKER': '1', 'DMLC_PS_ROOT_PORT': '9000', 'PATH': '/opt/amazon/bin:/usr/local/nvidia/bin:/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin:/opt/amazon/bin:/opt/amazon/bin', 'SAGEMAKER_HTTP_PORT': '8080', 'HOME': '/root', 'PYTHONUNBUFFERED': 'TRUE', 'CANONICAL_ENVROOT': '/opt/amazon', 'LD_LIBRARY_PATH': '/opt/amazon/lib/python2.7/site-packages/cv2/../../../../lib:/usr/local/nvidia/lib64:/opt/amazon/lib', 'LANG': 'en_US.utf8', 'DMLC_INTERFACE': 'eth0', 'SHLVL': '1', 'DMLC_PS_ROOT_URI': '10.2.95.208', 'AWS_REGION': 'us-east-1', 'NVIDIA_VISIBLE_DEVICES': 'void', 'TRAINING_JOB_NAME': 'pca-2020-02-18-02-03-36-037', 'PROTOCOL_BUFFERS_PYTHON_IMPLEMENTATION': 'cpp', 'ENVROOT': '/opt/amazon', 'SAGEMAKER_DATA_PATH': '/opt/ml', 'NVIDIA_DRIVER_CAPABILITIES': 'compute,utility', 'NVIDIA_REQUIRE_CUDA': 'cuda>=9.0', 'OMP_NUM_THREADS': '2', 'HOSTNAME': 'ip-10-2-95-208.ec2.internal', 'AWS_CONTAINER_CREDENTIALS_RELATIVE_URI': '/v2/credentials/642304cc-1ed4-4342-b0f6-b9c4f36bbe0a', 'DMLC_ROLE': 'scheduler', 'PWD': '/', 'DMLC_NUM_SERVER': '1', 'TRAINING_JOB_ARN': 'arn:aws:sagemaker:us-east-1:001178231653:training-job/pca-2020-02-18-02-03-36-037', 'AWS_EXECUTION_ENV': 'AWS_ECS_EC2'}\u001b[0m\n",
      "\u001b[34m[02/18/2020 02:06:18 INFO 140378775521088] Launching parameter server for role server\u001b[0m\n",
      "\u001b[34m[02/18/2020 02:06:18 INFO 140378775521088] {'ECS_CONTAINER_METADATA_URI': 'http://169.254.170.2/v3/47ec9ef1-5bf3-4241-847a-f87ce6e331de', 'PROTOCOL_BUFFERS_PYTHON_IMPLEMENTATION_VERSION': '2', 'PATH': '/opt/amazon/bin:/usr/local/nvidia/bin:/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin:/opt/amazon/bin:/opt/amazon/bin', 'SAGEMAKER_HTTP_PORT': '8080', 'HOME': '/root', 'PYTHONUNBUFFERED': 'TRUE', 'CANONICAL_ENVROOT': '/opt/amazon', 'LD_LIBRARY_PATH': '/opt/amazon/lib/python2.7/site-packages/cv2/../../../../lib:/usr/local/nvidia/lib64:/opt/amazon/lib', 'LANG': 'en_US.utf8', 'DMLC_INTERFACE': 'eth0', 'SHLVL': '1', 'AWS_REGION': 'us-east-1', 'NVIDIA_VISIBLE_DEVICES': 'void', 'TRAINING_JOB_NAME': 'pca-2020-02-18-02-03-36-037', 'PROTOCOL_BUFFERS_PYTHON_IMPLEMENTATION': 'cpp', 'ENVROOT': '/opt/amazon', 'SAGEMAKER_DATA_PATH': '/opt/ml', 'NVIDIA_DRIVER_CAPABILITIES': 'compute,utility', 'NVIDIA_REQUIRE_CUDA': 'cuda>=9.0', 'OMP_NUM_THREADS': '2', 'HOSTNAME': 'ip-10-2-95-208.ec2.internal', 'AWS_CONTAINER_CREDENTIALS_RELATIVE_URI': '/v2/credentials/642304cc-1ed4-4342-b0f6-b9c4f36bbe0a', 'PWD': '/', 'TRAINING_JOB_ARN': 'arn:aws:sagemaker:us-east-1:001178231653:training-job/pca-2020-02-18-02-03-36-037', 'AWS_EXECUTION_ENV': 'AWS_ECS_EC2'}\u001b[0m\n",
      "\u001b[34m[02/18/2020 02:06:18 INFO 140378775521088] envs={'ECS_CONTAINER_METADATA_URI': 'http://169.254.170.2/v3/47ec9ef1-5bf3-4241-847a-f87ce6e331de', 'PROTOCOL_BUFFERS_PYTHON_IMPLEMENTATION_VERSION': '2', 'DMLC_NUM_WORKER': '1', 'DMLC_PS_ROOT_PORT': '9000', 'PATH': '/opt/amazon/bin:/usr/local/nvidia/bin:/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin:/opt/amazon/bin:/opt/amazon/bin', 'SAGEMAKER_HTTP_PORT': '8080', 'HOME': '/root', 'PYTHONUNBUFFERED': 'TRUE', 'CANONICAL_ENVROOT': '/opt/amazon', 'LD_LIBRARY_PATH': '/opt/amazon/lib/python2.7/site-packages/cv2/../../../../lib:/usr/local/nvidia/lib64:/opt/amazon/lib', 'LANG': 'en_US.utf8', 'DMLC_INTERFACE': 'eth0', 'SHLVL': '1', 'DMLC_PS_ROOT_URI': '10.2.95.208', 'AWS_REGION': 'us-east-1', 'NVIDIA_VISIBLE_DEVICES': 'void', 'TRAINING_JOB_NAME': 'pca-2020-02-18-02-03-36-037', 'PROTOCOL_BUFFERS_PYTHON_IMPLEMENTATION': 'cpp', 'ENVROOT': '/opt/amazon', 'SAGEMAKER_DATA_PATH': '/opt/ml', 'NVIDIA_DRIVER_CAPABILITIES': 'compute,utility', 'NVIDIA_REQUIRE_CUDA': 'cuda>=9.0', 'OMP_NUM_THREADS': '2', 'HOSTNAME': 'ip-10-2-95-208.ec2.internal', 'AWS_CONTAINER_CREDENTIALS_RELATIVE_URI': '/v2/credentials/642304cc-1ed4-4342-b0f6-b9c4f36bbe0a', 'DMLC_ROLE': 'server', 'PWD': '/', 'DMLC_NUM_SERVER': '1', 'TRAINING_JOB_ARN': 'arn:aws:sagemaker:us-east-1:001178231653:training-job/pca-2020-02-18-02-03-36-037', 'AWS_EXECUTION_ENV': 'AWS_ECS_EC2'}\u001b[0m\n",
      "\u001b[34m[02/18/2020 02:06:18 INFO 140378775521088] Environment: {'ECS_CONTAINER_METADATA_URI': 'http://169.254.170.2/v3/47ec9ef1-5bf3-4241-847a-f87ce6e331de', 'PROTOCOL_BUFFERS_PYTHON_IMPLEMENTATION_VERSION': '2', 'DMLC_PS_ROOT_PORT': '9000', 'DMLC_NUM_WORKER': '1', 'SAGEMAKER_HTTP_PORT': '8080', 'PATH': '/opt/amazon/bin:/usr/local/nvidia/bin:/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin:/opt/amazon/bin:/opt/amazon/bin', 'PYTHONUNBUFFERED': 'TRUE', 'CANONICAL_ENVROOT': '/opt/amazon', 'LD_LIBRARY_PATH': '/opt/amazon/lib/python2.7/site-packages/cv2/../../../../lib:/usr/local/nvidia/lib64:/opt/amazon/lib', 'LANG': 'en_US.utf8', 'DMLC_INTERFACE': 'eth0', 'SHLVL': '1', 'DMLC_PS_ROOT_URI': '10.2.95.208', 'AWS_REGION': 'us-east-1', 'NVIDIA_VISIBLE_DEVICES': 'void', 'TRAINING_JOB_NAME': 'pca-2020-02-18-02-03-36-037', 'HOME': '/root', 'PROTOCOL_BUFFERS_PYTHON_IMPLEMENTATION': 'cpp', 'ENVROOT': '/opt/amazon', 'SAGEMAKER_DATA_PATH': '/opt/ml', 'NVIDIA_DRIVER_CAPABILITIES': 'compute,utility', 'NVIDIA_REQUIRE_CUDA': 'cuda>=9.0', 'OMP_NUM_THREADS': '2', 'HOSTNAME': 'ip-10-2-95-208.ec2.internal', 'AWS_CONTAINER_CREDENTIALS_RELATIVE_URI': '/v2/credentials/642304cc-1ed4-4342-b0f6-b9c4f36bbe0a', 'DMLC_ROLE': 'worker', 'PWD': '/', 'DMLC_NUM_SERVER': '1', 'TRAINING_JOB_ARN': 'arn:aws:sagemaker:us-east-1:001178231653:training-job/pca-2020-02-18-02-03-36-037', 'AWS_EXECUTION_ENV': 'AWS_ECS_EC2'}\u001b[0m\n",
      "\u001b[34mProcess 60 is a shell:scheduler.\u001b[0m\n",
      "\u001b[34mProcess 69 is a shell:server.\u001b[0m\n",
      "\u001b[34mProcess 1 is a worker.\u001b[0m\n",
      "\u001b[34m[02/18/2020 02:06:18 INFO 140378775521088] Using default worker.\u001b[0m\n",
      "\u001b[34m[02/18/2020 02:06:18 INFO 140378775521088] Loaded iterator creator application/x-recordio-protobuf for content type ('application/x-recordio-protobuf', '1.0')\u001b[0m\n",
      "\u001b[34m[02/18/2020 02:06:18 INFO 140378775521088] Loaded iterator creator application/x-labeled-vector-protobuf for content type ('application/x-labeled-vector-protobuf', '1.0')\u001b[0m\n",
      "\u001b[34m[02/18/2020 02:06:18 INFO 140378775521088] Loaded iterator creator protobuf for content type ('protobuf', '1.0')\u001b[0m\n",
      "\u001b[34m[02/18/2020 02:06:18 INFO 140378775521088] Create Store: dist_sync\u001b[0m\n",
      "\u001b[34m[02/18/2020 02:06:19 INFO 140378775521088] nvidia-smi took: 0.025171995163 secs to identify 0 gpus\u001b[0m\n",
      "\u001b[34m[02/18/2020 02:06:19 INFO 140378775521088] Number of GPUs being used: 0\u001b[0m\n",
      "\u001b[34m[02/18/2020 02:06:19 INFO 140378775521088] The default executor is <PCAExecutor on cpu(0)>.\u001b[0m\n",
      "\u001b[34m[02/18/2020 02:06:19 INFO 140378775521088] 20 feature(s) found in 'data'.\u001b[0m\n",
      "\u001b[34m[02/18/2020 02:06:19 INFO 140378775521088] <PCAExecutor on cpu(0)> is assigned to batch slice from 0 to 499.\u001b[0m\n",
      "\u001b[34m#metrics {\"Metrics\": {\"initialize.time\": {\"count\": 1, \"max\": 585.4029655456543, \"sum\": 585.4029655456543, \"min\": 585.4029655456543}}, \"EndTime\": 1581991579.355124, \"Dimensions\": {\"Host\": \"algo-1\", \"Operation\": \"training\", \"Algorithm\": \"PCA\"}, \"StartTime\": 1581991578.765684}\n",
      "\u001b[0m\n",
      "\u001b[34m#metrics {\"Metrics\": {\"Max Batches Seen Between Resets\": {\"count\": 1, \"max\": 0, \"sum\": 0.0, \"min\": 0}, \"Number of Batches Since Last Reset\": {\"count\": 1, \"max\": 0, \"sum\": 0.0, \"min\": 0}, \"Number of Records Since Last Reset\": {\"count\": 1, \"max\": 0, \"sum\": 0.0, \"min\": 0}, \"Total Batches Seen\": {\"count\": 1, \"max\": 0, \"sum\": 0.0, \"min\": 0}, \"Total Records Seen\": {\"count\": 1, \"max\": 0, \"sum\": 0.0, \"min\": 0}, \"Max Records Seen Between Resets\": {\"count\": 1, \"max\": 0, \"sum\": 0.0, \"min\": 0}, \"Reset Count\": {\"count\": 1, \"max\": 0, \"sum\": 0.0, \"min\": 0}}, \"EndTime\": 1581991579.355368, \"Dimensions\": {\"Host\": \"algo-1\", \"Meta\": \"init_train_data_iter\", \"Operation\": \"training\", \"Algorithm\": \"PCA\"}, \"StartTime\": 1581991579.35531}\n",
      "\u001b[0m\n",
      "\u001b[34m[2020-02-18 02:06:19.355] [tensorio] [info] epoch_stats={\"data_pipeline\": \"/opt/ml/input/data/train\", \"epoch\": 0, \"duration\": 589, \"num_examples\": 1, \"num_bytes\": 52000}\u001b[0m\n",
      "\u001b[34m[2020-02-18 02:06:19.401] [tensorio] [info] epoch_stats={\"data_pipeline\": \"/opt/ml/input/data/train\", \"epoch\": 1, \"duration\": 36, \"num_examples\": 2, \"num_bytes\": 76960}\u001b[0m\n",
      "\u001b[34m#metrics {\"Metrics\": {\"epochs\": {\"count\": 1, \"max\": 1, \"sum\": 1.0, \"min\": 1}, \"update.time\": {\"count\": 1, \"max\": 46.23913764953613, \"sum\": 46.23913764953613, \"min\": 46.23913764953613}}, \"EndTime\": 1581991579.401924, \"Dimensions\": {\"Host\": \"algo-1\", \"Operation\": \"training\", \"Algorithm\": \"PCA\"}, \"StartTime\": 1581991579.355227}\n",
      "\u001b[0m\n",
      "\u001b[34m[02/18/2020 02:06:19 INFO 140378775521088] #progress_metric: host=algo-1, completed 100 % of epochs\u001b[0m\n",
      "\u001b[34m#metrics {\"Metrics\": {\"Max Batches Seen Between Resets\": {\"count\": 1, \"max\": 2, \"sum\": 2.0, \"min\": 2}, \"Number of Batches Since Last Reset\": {\"count\": 1, \"max\": 2, \"sum\": 2.0, \"min\": 2}, \"Number of Records Since Last Reset\": {\"count\": 1, \"max\": 740, \"sum\": 740.0, \"min\": 740}, \"Total Batches Seen\": {\"count\": 1, \"max\": 2, \"sum\": 2.0, \"min\": 2}, \"Total Records Seen\": {\"count\": 1, \"max\": 740, \"sum\": 740.0, \"min\": 740}, \"Max Records Seen Between Resets\": {\"count\": 1, \"max\": 740, \"sum\": 740.0, \"min\": 740}, \"Reset Count\": {\"count\": 1, \"max\": 1, \"sum\": 1.0, \"min\": 1}}, \"EndTime\": 1581991579.40221, \"Dimensions\": {\"Host\": \"algo-1\", \"Meta\": \"training_data_iter\", \"Operation\": \"training\", \"Algorithm\": \"PCA\", \"epoch\": 0}, \"StartTime\": 1581991579.355654}\n",
      "\u001b[0m\n",
      "\u001b[34m[02/18/2020 02:06:19 INFO 140378775521088] #throughput_metric: host=algo-1, train throughput=15857.4010249 records/second\u001b[0m\n",
      "\u001b[34m#metrics {\"Metrics\": {\"finalize.time\": {\"count\": 1, \"max\": 20.998001098632812, \"sum\": 20.998001098632812, \"min\": 20.998001098632812}}, \"EndTime\": 1581991579.423485, \"Dimensions\": {\"Host\": \"algo-1\", \"Operation\": \"training\", \"Algorithm\": \"PCA\"}, \"StartTime\": 1581991579.402002}\n",
      "\u001b[0m\n",
      "\u001b[34m[02/18/2020 02:06:19 INFO 140378775521088] Test data is not provided.\u001b[0m\n",
      "\u001b[34m#metrics {\"Metrics\": {\"totaltime\": {\"count\": 1, \"max\": 816.9910907745361, \"sum\": 816.9910907745361, \"min\": 816.9910907745361}, \"setuptime\": {\"count\": 1, \"max\": 39.095163345336914, \"sum\": 39.095163345336914, \"min\": 39.095163345336914}}, \"EndTime\": 1581991579.430532, \"Dimensions\": {\"Host\": \"algo-1\", \"Operation\": \"training\", \"Algorithm\": \"PCA\"}, \"StartTime\": 1581991579.423543}\n",
      "\u001b[0m\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "2020-02-18 02:06:27 Uploading - Uploading generated training model\n",
      "2020-02-18 02:06:27 Completed - Training job completed\n",
      "Training seconds: 46\n",
      "Billable seconds: 46\n",
      "CPU times: user 415 ms, sys: 22.9 ms, total: 438 ms\n",
      "Wall time: 3min 12s\n"
     ]
    }
   ],
   "source": [
    "%%time\n",
    "pca_SM.fit(pca_SM.record_set(train_data))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0"
      ]
     },
     "execution_count": 15,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Store model artifacts on S3\n",
    "job_name = pca_SM.latest_training_job.name\n",
    "model_key = \"absence/\" + job_name + \"/output/model.tar.gz\"\n",
    "\n",
    "boto3.resource('s3').Bucket(bucket).download_file(model_key, 'model.tar.gz')\n",
    "os.system('tar -zxvf model.tar.gz')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [],
   "source": [
    "# After the model is decompressed, we can load the ND array using MXNet\n",
    "import mxnet as mx\n",
    "pca_model_params = mx.ndarray.load('model_algo-1')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Exact explained-variance-ratio vector\n",
    "s=pd.DataFrame(pca_model_params['s'].asnumpy())\n",
    "v=pd.DataFrame(pca_model_params['v'].asnumpy())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0    0.622284\n",
       "dtype: float32"
      ]
     },
     "execution_count": 18,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Calculate the variance explained by the largest n components that we want to keep. \n",
    "# Take the top 5 components\n",
    "# The largest 5 components explain ~62% of the total variance in our dataset\n",
    "s.iloc[14:,:].apply(lambda x: x*x).sum()/s.apply(lambda x: x*x).sum()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [],
   "source": [
    "# After keeping the top 5 components take only the 5 largest components from original s and v matrix\n",
    "s_5=s.iloc[14:,:]\n",
    "v_5=v.iloc[:,14:]\n",
    "v_5.columns=[0,1,2,3,4]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "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>0</th>\n",
       "      <th>1</th>\n",
       "      <th>2</th>\n",
       "      <th>3</th>\n",
       "      <th>4</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0.407281</td>\n",
       "      <td>-0.448665</td>\n",
       "      <td>0.193074</td>\n",
       "      <td>0.289915</td>\n",
       "      <td>-0.059501</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>0.217295</td>\n",
       "      <td>-0.079499</td>\n",
       "      <td>-0.023497</td>\n",
       "      <td>-0.452234</td>\n",
       "      <td>-0.036570</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>-0.584830</td>\n",
       "      <td>-0.525670</td>\n",
       "      <td>0.547078</td>\n",
       "      <td>-0.047001</td>\n",
       "      <td>-0.035020</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>0.160280</td>\n",
       "      <td>-0.319543</td>\n",
       "      <td>-0.043324</td>\n",
       "      <td>-0.762302</td>\n",
       "      <td>0.045018</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>0.098019</td>\n",
       "      <td>0.332768</td>\n",
       "      <td>0.286093</td>\n",
       "      <td>-0.152991</td>\n",
       "      <td>-0.065728</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "          0         1         2         3         4\n",
       "0  0.407281 -0.448665  0.193074  0.289915 -0.059501\n",
       "1  0.217295 -0.079499 -0.023497 -0.452234 -0.036570\n",
       "2 -0.584830 -0.525670  0.547078 -0.047001 -0.035020\n",
       "3  0.160280 -0.319543 -0.043324 -0.762302  0.045018\n",
       "4  0.098019  0.332768  0.286093 -0.152991 -0.065728"
      ]
     },
     "execution_count": 20,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "v_5.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Examine the makeup of one of the PCA components based on the weightings of the original features \n",
    "# that are included in the component\n",
    "# Understand what the key positive and negative attributes are for the component\n",
    "component_num=1\n",
    "\n",
    "first_comp = v_5[5-component_num]\n",
    "comps = pd.DataFrame(list(zip(first_comp, absence_scaled.columns)), columns=['weights', 'features'])\n",
    "comps['abs_weights']=comps['weights'].apply(lambda x: np.abs(x))\n",
    "ax=sns.barplot(data=comps.sort_values('abs_weights', ascending=False).head(10), x=\"weights\", y=\"features\", palette=\"Blues_d\")\n",
    "ax.set_title(\"PCA Component Makeup: #\" + str(component_num))\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "-----------------------!CPU times: user 354 ms, sys: 29.7 ms, total: 384 ms\n",
      "Wall time: 11min 32s\n"
     ]
    }
   ],
   "source": [
    "%%time\n",
    "pca_predictor = pca_SM.deploy(initial_instance_count=1, \n",
    "                                 instance_type='ml.t2.medium')\n",
    "# Deploy the PCA model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "CPU times: user 52.7 ms, sys: 0 ns, total: 52.7 ms\n",
      "Wall time: 259 ms\n"
     ]
    }
   ],
   "source": [
    "%%time\n",
    "result = pca_predictor.predict(train_data)\n",
    "# Pass our original dataset to the model so that we can transform the data using the model we created\n",
    "# Take the largest 5 components and this will reduce the dimensionality of our data from 19 to 5"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Create a dataset where each absenteeism is described by the 5 principle components\n",
    "PCA_list=['Social drinker', 'Distance from Residence to Work', 'Weight', 'Education', 'Body mass index']\n",
    "absence_transformed=pd.DataFrame()\n",
    "for a in result:\n",
    "    b=a.label['projection'].float32_tensor.values\n",
    "    absence_transformed=absence_transformed.append([list(b)])\n",
    "absence_transformed.index=absence_scaled.index\n",
    "absence_transformed=absence_transformed.iloc[:,14:]\n",
    "absence_transformed.columns=PCA_list"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "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>Social drinker</th>\n",
       "      <th>Distance from Residence to Work</th>\n",
       "      <th>Weight</th>\n",
       "      <th>Education</th>\n",
       "      <th>Body mass index</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Reason for absence</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>26</th>\n",
       "      <td>0.220163</td>\n",
       "      <td>0.277262</td>\n",
       "      <td>0.023891</td>\n",
       "      <td>0.416488</td>\n",
       "      <td>-0.574165</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>-0.813139</td>\n",
       "      <td>0.473745</td>\n",
       "      <td>-0.941796</td>\n",
       "      <td>0.071819</td>\n",
       "      <td>-0.446411</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>23</th>\n",
       "      <td>0.119235</td>\n",
       "      <td>-0.130268</td>\n",
       "      <td>-0.052720</td>\n",
       "      <td>0.539152</td>\n",
       "      <td>-0.643213</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>-0.839607</td>\n",
       "      <td>0.590692</td>\n",
       "      <td>0.207264</td>\n",
       "      <td>0.134280</td>\n",
       "      <td>-0.147293</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>23</th>\n",
       "      <td>-0.115692</td>\n",
       "      <td>0.061643</td>\n",
       "      <td>0.277216</td>\n",
       "      <td>0.362107</td>\n",
       "      <td>-0.585175</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                    Social drinker  Distance from Residence to Work    Weight  \\\n",
       "Reason for absence                                                              \n",
       "26                        0.220163                         0.277262  0.023891   \n",
       "0                        -0.813139                         0.473745 -0.941796   \n",
       "23                        0.119235                        -0.130268 -0.052720   \n",
       "7                        -0.839607                         0.590692  0.207264   \n",
       "23                       -0.115692                         0.061643  0.277216   \n",
       "\n",
       "                    Education  Body mass index  \n",
       "Reason for absence                              \n",
       "26                   0.416488        -0.574165  \n",
       "0                    0.071819        -0.446411  \n",
       "23                   0.539152        -0.643213  \n",
       "7                    0.134280        -0.147293  \n",
       "23                   0.362107        -0.585175  "
      ]
     },
     "execution_count": 25,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "absence_transformed.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Use the Kmeans algorithm to segment the absenteeism by the 5 PCA attributes we have created.\n",
    "train_data = absence_transformed.values.astype('float32')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Find the top 7 clusters from our dataset\n",
    "from sagemaker import KMeans\n",
    "\n",
    "num_clusters = 7\n",
    "kmeans = KMeans(role=role,\n",
    "                train_instance_count=1,\n",
    "                train_instance_type='ml.c4.xlarge',\n",
    "                output_path='s3://'+ bucket +'/absence/',              \n",
    "                k=num_clusters)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "2020-02-18 02:25:53 Starting - Starting the training job...\n",
      "2020-02-18 02:25:55 Starting - Launching requested ML instances......\n",
      "2020-02-18 02:26:59 Starting - Preparing the instances for training...\n",
      "2020-02-18 02:27:43 Downloading - Downloading input data...\n",
      "2020-02-18 02:27:57 Training - Downloading the training image......\n",
      "2020-02-18 02:29:22 Uploading - Uploading generated training model\n",
      "2020-02-18 02:29:22 Completed - Training job completed\n",
      "\u001b[34mDocker entrypoint called with argument(s): train\u001b[0m\n",
      "\u001b[34m[02/18/2020 02:29:13 INFO 140491501401920] Reading default configuration from /opt/amazon/lib/python2.7/site-packages/algorithm/resources/default-input.json: {u'_enable_profiler': u'false', u'_tuning_objective_metric': u'', u'_num_gpus': u'auto', u'local_lloyd_num_trials': u'auto', u'_log_level': u'info', u'_kvstore': u'auto', u'local_lloyd_init_method': u'kmeans++', u'force_dense': u'true', u'epochs': u'1', u'init_method': u'random', u'local_lloyd_tol': u'0.0001', u'local_lloyd_max_iter': u'300', u'_disable_wait_to_read': u'false', u'extra_center_factor': u'auto', u'eval_metrics': u'[\"msd\"]', u'_num_kv_servers': u'1', u'mini_batch_size': u'5000', u'half_life_time_size': u'0', u'_num_slices': u'1'}\u001b[0m\n",
      "\u001b[34m[02/18/2020 02:29:13 INFO 140491501401920] Reading provided configuration from /opt/ml/input/config/hyperparameters.json: {u'feature_dim': u'5', u'k': u'7', u'force_dense': u'True'}\u001b[0m\n",
      "\u001b[34m[02/18/2020 02:29:13 INFO 140491501401920] Final configuration: {u'_tuning_objective_metric': u'', u'extra_center_factor': u'auto', u'local_lloyd_init_method': u'kmeans++', u'force_dense': u'True', u'epochs': u'1', u'feature_dim': u'5', u'local_lloyd_tol': u'0.0001', u'_disable_wait_to_read': u'false', u'eval_metrics': u'[\"msd\"]', u'_num_kv_servers': u'1', u'mini_batch_size': u'5000', u'_enable_profiler': u'false', u'_num_gpus': u'auto', u'local_lloyd_num_trials': u'auto', u'_log_level': u'info', u'init_method': u'random', u'half_life_time_size': u'0', u'local_lloyd_max_iter': u'300', u'_kvstore': u'auto', u'k': u'7', u'_num_slices': u'1'}\u001b[0m\n",
      "\u001b[34m[02/18/2020 02:29:13 WARNING 140491501401920] Loggers have already been setup.\u001b[0m\n",
      "\u001b[34mProcess 1 is a worker.\u001b[0m\n",
      "\u001b[34m[02/18/2020 02:29:13 INFO 140491501401920] Using default worker.\u001b[0m\n",
      "\u001b[34m[02/18/2020 02:29:13 INFO 140491501401920] Loaded iterator creator application/x-recordio-protobuf for content type ('application/x-recordio-protobuf', '1.0')\u001b[0m\n",
      "\u001b[34m[02/18/2020 02:29:13 INFO 140491501401920] Create Store: local\u001b[0m\n",
      "\u001b[34m[02/18/2020 02:29:13 INFO 140491501401920] nvidia-smi took: 0.0251748561859 secs to identify 0 gpus\u001b[0m\n",
      "\u001b[34m[02/18/2020 02:29:13 INFO 140491501401920] Number of GPUs being used: 0\u001b[0m\n",
      "\u001b[34m[02/18/2020 02:29:13 INFO 140491501401920] Setting up with params: {u'_tuning_objective_metric': u'', u'extra_center_factor': u'auto', u'local_lloyd_init_method': u'kmeans++', u'force_dense': u'True', u'epochs': u'1', u'feature_dim': u'5', u'local_lloyd_tol': u'0.0001', u'_disable_wait_to_read': u'false', u'eval_metrics': u'[\"msd\"]', u'_num_kv_servers': u'1', u'mini_batch_size': u'5000', u'_enable_profiler': u'false', u'_num_gpus': u'auto', u'local_lloyd_num_trials': u'auto', u'_log_level': u'info', u'init_method': u'random', u'half_life_time_size': u'0', u'local_lloyd_max_iter': u'300', u'_kvstore': u'auto', u'k': u'7', u'_num_slices': u'1'}\u001b[0m\n",
      "\u001b[34m[02/18/2020 02:29:13 INFO 140491501401920] 'extra_center_factor' was set to 'auto', evaluated to 10.\u001b[0m\n",
      "\u001b[34m[02/18/2020 02:29:13 INFO 140491501401920] Number of GPUs being used: 0\u001b[0m\n",
      "\u001b[34m[02/18/2020 02:29:13 INFO 140491501401920] number of center slices 1\u001b[0m\n",
      "\u001b[34m[02/18/2020 02:29:13 WARNING 140491501401920] Batch size 5000 is bigger than the first batch data. Effective batch size used to initialize is 740\u001b[0m\n",
      "\u001b[34m#metrics {\"Metrics\": {\"Max Batches Seen Between Resets\": {\"count\": 1, \"max\": 1, \"sum\": 1.0, \"min\": 1}, \"Number of Batches Since Last Reset\": {\"count\": 1, \"max\": 1, \"sum\": 1.0, \"min\": 1}, \"Number of Records Since Last Reset\": {\"count\": 1, \"max\": 740, \"sum\": 740.0, \"min\": 740}, \"Total Batches Seen\": {\"count\": 1, \"max\": 1, \"sum\": 1.0, \"min\": 1}, \"Total Records Seen\": {\"count\": 1, \"max\": 740, \"sum\": 740.0, \"min\": 740}, \"Max Records Seen Between Resets\": {\"count\": 1, \"max\": 740, \"sum\": 740.0, \"min\": 740}, \"Reset Count\": {\"count\": 1, \"max\": 0, \"sum\": 0.0, \"min\": 0}}, \"EndTime\": 1581992953.656437, \"Dimensions\": {\"Host\": \"algo-1\", \"Meta\": \"init_train_data_iter\", \"Operation\": \"training\", \"Algorithm\": \"AWS/KMeansWebscale\"}, \"StartTime\": 1581992953.656406}\n",
      "\u001b[0m\n",
      "\u001b[34m[2020-02-18 02:29:13.656] [tensorio] [info] epoch_stats={\"data_pipeline\": \"/opt/ml/input/data/train\", \"epoch\": 0, \"duration\": 30, \"num_examples\": 1, \"num_bytes\": 32560}\u001b[0m\n",
      "\u001b[34m[2020-02-18 02:29:13.699] [tensorio] [info] epoch_stats={\"data_pipeline\": \"/opt/ml/input/data/train\", \"epoch\": 1, \"duration\": 42, \"num_examples\": 1, \"num_bytes\": 32560}\u001b[0m\n",
      "\u001b[34m[02/18/2020 02:29:13 INFO 140491501401920] processed a total of 740 examples\u001b[0m\n",
      "\u001b[34m[02/18/2020 02:29:13 INFO 140491501401920] #progress_metric: host=algo-1, completed 100 % of epochs\u001b[0m\n",
      "\u001b[34m#metrics {\"Metrics\": {\"Max Batches Seen Between Resets\": {\"count\": 1, \"max\": 1, \"sum\": 1.0, \"min\": 1}, \"Number of Batches Since Last Reset\": {\"count\": 1, \"max\": 1, \"sum\": 1.0, \"min\": 1}, \"Number of Records Since Last Reset\": {\"count\": 1, \"max\": 740, \"sum\": 740.0, \"min\": 740}, \"Total Batches Seen\": {\"count\": 1, \"max\": 2, \"sum\": 2.0, \"min\": 2}, \"Total Records Seen\": {\"count\": 1, \"max\": 1480, \"sum\": 1480.0, \"min\": 1480}, \"Max Records Seen Between Resets\": {\"count\": 1, \"max\": 740, \"sum\": 740.0, \"min\": 740}, \"Reset Count\": {\"count\": 1, \"max\": 1, \"sum\": 1.0, \"min\": 1}}, \"EndTime\": 1581992953.700207, \"Dimensions\": {\"Host\": \"algo-1\", \"Meta\": \"training_data_iter\", \"Operation\": \"training\", \"Algorithm\": \"AWS/KMeansWebscale\", \"epoch\": 0}, \"StartTime\": 1581992953.656665}\n",
      "\u001b[0m\n",
      "\u001b[34m[02/18/2020 02:29:13 INFO 140491501401920] #throughput_metric: host=algo-1, train throughput=16946.7753578 records/second\u001b[0m\n",
      "\u001b[34m[02/18/2020 02:29:13 WARNING 140491501401920] wait_for_all_workers will not sync workers since the kv store is not running distributed\u001b[0m\n",
      "\u001b[34m[02/18/2020 02:29:13 INFO 140491501401920] shrinking 70 centers into 7\u001b[0m\n",
      "\u001b[34m[02/18/2020 02:29:13 INFO 140491501401920] local kmeans attempt #0. Current mean square distance 0.267164\u001b[0m\n",
      "\u001b[34m[02/18/2020 02:29:13 INFO 140491501401920] local kmeans attempt #1. Current mean square distance 0.255519\u001b[0m\n",
      "\u001b[34m[02/18/2020 02:29:13 INFO 140491501401920] local kmeans attempt #2. Current mean square distance 0.268737\u001b[0m\n",
      "\u001b[34m[02/18/2020 02:29:13 INFO 140491501401920] local kmeans attempt #3. Current mean square distance 0.254238\u001b[0m\n",
      "\u001b[34m[02/18/2020 02:29:13 INFO 140491501401920] local kmeans attempt #4. Current mean square distance 0.272609\u001b[0m\n",
      "\u001b[34m[02/18/2020 02:29:13 INFO 140491501401920] local kmeans attempt #5. Current mean square distance 0.278503\u001b[0m\n",
      "\u001b[34m[02/18/2020 02:29:13 INFO 140491501401920] local kmeans attempt #6. Current mean square distance 0.268334\u001b[0m\n",
      "\u001b[34m[02/18/2020 02:29:13 INFO 140491501401920] local kmeans attempt #7. Current mean square distance 0.242600\u001b[0m\n",
      "\u001b[34m[02/18/2020 02:29:13 INFO 140491501401920] local kmeans attempt #8. Current mean square distance 0.272297\u001b[0m\n",
      "\u001b[34m[02/18/2020 02:29:13 INFO 140491501401920] local kmeans attempt #9. Current mean square distance 0.269440\u001b[0m\n",
      "\u001b[34m[02/18/2020 02:29:13 INFO 140491501401920] finished shrinking process. Mean Square Distance = 0\u001b[0m\n",
      "\u001b[34m[02/18/2020 02:29:13 INFO 140491501401920] #quality_metric: host=algo-1, train msd <loss>=0.242599859834\u001b[0m\n",
      "\u001b[34m[02/18/2020 02:29:13 INFO 140491501401920] compute all data-center distances: inner product took: 34.1606%, (0.016924 secs)\u001b[0m\n",
      "\u001b[34m[02/18/2020 02:29:13 INFO 140491501401920] collect from kv store took: 13.9294%, (0.006901 secs)\u001b[0m\n",
      "\u001b[34m[02/18/2020 02:29:13 INFO 140491501401920] splitting centers key-value pair took: 13.7335%, (0.006804 secs)\u001b[0m\n",
      "\u001b[34m[02/18/2020 02:29:13 INFO 140491501401920] gradient: one_hot took: 10.3346%, (0.005120 secs)\u001b[0m\n",
      "\u001b[34m[02/18/2020 02:29:13 INFO 140491501401920] gradient: cluster center took: 9.6137%, (0.004763 secs)\u001b[0m\n",
      "\u001b[34m[02/18/2020 02:29:13 INFO 140491501401920] compute all data-center distances: point norm took: 8.0415%, (0.003984 secs)\u001b[0m\n",
      "\u001b[34m[02/18/2020 02:29:13 INFO 140491501401920] predict compute msd took: 4.1319%, (0.002047 secs)\u001b[0m\n",
      "\u001b[34m[02/18/2020 02:29:13 INFO 140491501401920] batch data loading with context took: 3.4538%, (0.001711 secs)\u001b[0m\n",
      "\u001b[34m[02/18/2020 02:29:13 INFO 140491501401920] update state and report convergance took: 1.0212%, (0.000506 secs)\u001b[0m\n",
      "\u001b[34m[02/18/2020 02:29:13 INFO 140491501401920] gradient: cluster size  took: 0.8253%, (0.000409 secs)\u001b[0m\n",
      "\u001b[34m[02/18/2020 02:29:13 INFO 140491501401920] update set-up time took: 0.4158%, (0.000206 secs)\u001b[0m\n",
      "\u001b[34m[02/18/2020 02:29:13 INFO 140491501401920] compute all data-center distances: center norm took: 0.2964%, (0.000147 secs)\u001b[0m\n",
      "\u001b[34m[02/18/2020 02:29:13 INFO 140491501401920] predict minus dist took: 0.0423%, (0.000021 secs)\u001b[0m\n",
      "\u001b[34m[02/18/2020 02:29:13 INFO 140491501401920] TOTAL took: 0.0495429039001\u001b[0m\n",
      "\u001b[34m[02/18/2020 02:29:13 INFO 140491501401920] Number of GPUs being used: 0\u001b[0m\n",
      "\u001b[34m#metrics {\"Metrics\": {\"finalize.time\": {\"count\": 1, \"max\": 276.57604217529297, \"sum\": 276.57604217529297, \"min\": 276.57604217529297}, \"initialize.time\": {\"count\": 1, \"max\": 28.434038162231445, \"sum\": 28.434038162231445, \"min\": 28.434038162231445}, \"model.serialize.time\": {\"count\": 1, \"max\": 0.16188621520996094, \"sum\": 0.16188621520996094, \"min\": 0.16188621520996094}, \"update.time\": {\"count\": 1, \"max\": 43.35594177246094, \"sum\": 43.35594177246094, \"min\": 43.35594177246094}, \"epochs\": {\"count\": 1, \"max\": 1, \"sum\": 1.0, \"min\": 1}, \"state.serialize.time\": {\"count\": 1, \"max\": 1.177072525024414, \"sum\": 1.177072525024414, \"min\": 1.177072525024414}, \"_shrink.time\": {\"count\": 1, \"max\": 274.72710609436035, \"sum\": 274.72710609436035, \"min\": 274.72710609436035}}, \"EndTime\": 1581992953.9786, \"Dimensions\": {\"Host\": \"algo-1\", \"Operation\": \"training\", \"Algorithm\": \"AWS/KMeansWebscale\"}, \"StartTime\": 1581992953.625163}\n",
      "\u001b[0m\n",
      "\u001b[34m[02/18/2020 02:29:13 INFO 140491501401920] Test data is not provided.\u001b[0m\n",
      "\u001b[34m#metrics {\"Metrics\": {\"totaltime\": {\"count\": 1, \"max\": 419.1310405731201, \"sum\": 419.1310405731201, \"min\": 419.1310405731201}, \"setuptime\": {\"count\": 1, \"max\": 13.462066650390625, \"sum\": 13.462066650390625, \"min\": 13.462066650390625}}, \"EndTime\": 1581992953.978951, \"Dimensions\": {\"Host\": \"algo-1\", \"Operation\": \"training\", \"Algorithm\": \"AWS/KMeansWebscale\"}, \"StartTime\": 1581992953.978695}\n",
      "\u001b[0m\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Training seconds: 99\n",
      "Billable seconds: 99\n",
      "CPU times: user 552 ms, sys: 33.4 ms, total: 586 ms\n",
      "Wall time: 3min 42s\n"
     ]
    }
   ],
   "source": [
    "%%time\n",
    "kmeans.fit(kmeans.record_set(train_data))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "-----------------------!CPU times: user 358 ms, sys: 30 ms, total: 388 ms\n",
      "Wall time: 11min 32s\n"
     ]
    }
   ],
   "source": [
    "%%time\n",
    "kmeans_predictor = kmeans.deploy(initial_instance_count=1, \n",
    "                                 instance_type='ml.t2.medium')\n",
    "# Deploy the model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "CPU times: user 51.9 ms, sys: 4.45 ms, total: 56.3 ms\n",
      "Wall time: 324 ms\n"
     ]
    }
   ],
   "source": [
    "%%time\n",
    "result=kmeans_predictor.predict(train_data)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Breakdown of cluster counts and the distribution of clusters\n",
    "cluster_labels = [r.label['closest_cluster'].float32_tensor.values[0] for r in result]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "2.0    138\n",
       "0.0    114\n",
       "1.0    111\n",
       "4.0    110\n",
       "3.0     96\n",
       "5.0     89\n",
       "6.0     82\n",
       "Name: 0, dtype: int64"
      ]
     },
     "execution_count": 32,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "pd.DataFrame(cluster_labels)[0].value_counts()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/ec2-user/anaconda3/envs/python3/lib/python3.6/site-packages/scipy/stats/stats.py:1713: FutureWarning: Using a non-tuple sequence for multidimensional indexing is deprecated; use `arr[tuple(seq)]` instead of `arr[seq]`. In the future this will be interpreted as an array index, `arr[np.array(seq)]`, which will result either in an error or a different result.\n",
      "  return np.add.reduce(sorted[indexer] * weights, axis=axis) / sumval\n",
      "/home/ec2-user/anaconda3/envs/python3/lib/python3.6/site-packages/matplotlib/axes/_axes.py:6521: MatplotlibDeprecationWarning: \n",
      "The 'normed' kwarg was deprecated in Matplotlib 2.1 and will be removed in 3.1. Use 'density' instead.\n",
      "  alternative=\"'density'\", removal=\"3.1\")\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x216 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "ax=plt.subplots(figsize=(6,3))\n",
    "ax=sns.distplot(cluster_labels, kde=False)\n",
    "title=\"Histogram of Cluster Counts\"\n",
    "ax.set_title(title, fontsize=12)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0"
      ]
     },
     "execution_count": 34,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Access the underlying model to get the cluster centers\n",
    "# Centers will help describe which features characterize each cluster\n",
    "# First retrieve the K-Means model attributes\n",
    "job_name = kmeans.latest_training_job.name\n",
    "model_key = \"absence/\" + job_name + \"/output/model.tar.gz\"\n",
    "\n",
    "boto3.resource('s3').Bucket(bucket).download_file(model_key, 'model.tar.gz')\n",
    "os.system('tar -zxvf model.tar.gz')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {},
   "outputs": [],
   "source": [
    "Kmeans_model_params = mx.ndarray.load('model_algo-1')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Cluster Centroid Locations: The location of the centers of each cluster identified by the Kmeans algorithm\n",
    "cluster_centroids=pd.DataFrame(Kmeans_model_params[0].asnumpy())\n",
    "cluster_centroids.columns=absence_transformed.columns"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "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>Social drinker</th>\n",
       "      <th>Distance from Residence to Work</th>\n",
       "      <th>Weight</th>\n",
       "      <th>Education</th>\n",
       "      <th>Body mass index</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>-0.194499</td>\n",
       "      <td>-0.061751</td>\n",
       "      <td>0.357695</td>\n",
       "      <td>0.185599</td>\n",
       "      <td>-0.471168</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>0.072919</td>\n",
       "      <td>0.208198</td>\n",
       "      <td>-0.263245</td>\n",
       "      <td>0.256122</td>\n",
       "      <td>0.633057</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>0.170031</td>\n",
       "      <td>0.178539</td>\n",
       "      <td>-0.110450</td>\n",
       "      <td>0.290928</td>\n",
       "      <td>-0.498988</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>0.117853</td>\n",
       "      <td>-0.079949</td>\n",
       "      <td>-0.034841</td>\n",
       "      <td>-0.466216</td>\n",
       "      <td>0.614537</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>-0.145866</td>\n",
       "      <td>-0.142386</td>\n",
       "      <td>0.292171</td>\n",
       "      <td>0.223818</td>\n",
       "      <td>0.644043</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>0.222085</td>\n",
       "      <td>-0.052123</td>\n",
       "      <td>0.152401</td>\n",
       "      <td>-0.581722</td>\n",
       "      <td>-0.477426</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>-0.341128</td>\n",
       "      <td>-0.128244</td>\n",
       "      <td>-0.547628</td>\n",
       "      <td>-0.137222</td>\n",
       "      <td>-0.414144</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   Social drinker  Distance from Residence to Work    Weight  Education  \\\n",
       "0       -0.194499                        -0.061751  0.357695   0.185599   \n",
       "1        0.072919                         0.208198 -0.263245   0.256122   \n",
       "2        0.170031                         0.178539 -0.110450   0.290928   \n",
       "3        0.117853                        -0.079949 -0.034841  -0.466216   \n",
       "4       -0.145866                        -0.142386  0.292171   0.223818   \n",
       "5        0.222085                        -0.052123  0.152401  -0.581722   \n",
       "6       -0.341128                        -0.128244 -0.547628  -0.137222   \n",
       "\n",
       "   Body mass index  \n",
       "0        -0.471168  \n",
       "1         0.633057  \n",
       "2        -0.498988  \n",
       "3         0.614537  \n",
       "4         0.644043  \n",
       "5        -0.477426  \n",
       "6        -0.414144  "
      ]
     },
     "execution_count": 37,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "cluster_centroids"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1152x432 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Plot a heatmap of the centroids and their location in the transformed feature space\n",
    "plt.figure(figsize = (16, 6))\n",
    "ax = sns.heatmap(cluster_centroids.T, cmap = 'YlGnBu')\n",
    "ax.set_xlabel(\"Cluster\")\n",
    "plt.yticks(fontsize = 16)\n",
    "plt.xticks(fontsize = 16)\n",
    "ax.set_title(\"Attribute Value by Centroid\")\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "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>Social drinker</th>\n",
       "      <th>Distance from Residence to Work</th>\n",
       "      <th>Weight</th>\n",
       "      <th>Education</th>\n",
       "      <th>Body mass index</th>\n",
       "      <th>labels</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Reason for absence</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>26</th>\n",
       "      <td>0.220163</td>\n",
       "      <td>0.277262</td>\n",
       "      <td>0.023891</td>\n",
       "      <td>0.416488</td>\n",
       "      <td>-0.574165</td>\n",
       "      <td>2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>-0.813139</td>\n",
       "      <td>0.473745</td>\n",
       "      <td>-0.941796</td>\n",
       "      <td>0.071819</td>\n",
       "      <td>-0.446411</td>\n",
       "      <td>6</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>23</th>\n",
       "      <td>0.119235</td>\n",
       "      <td>-0.130268</td>\n",
       "      <td>-0.052720</td>\n",
       "      <td>0.539152</td>\n",
       "      <td>-0.643213</td>\n",
       "      <td>2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>-0.839607</td>\n",
       "      <td>0.590692</td>\n",
       "      <td>0.207264</td>\n",
       "      <td>0.134280</td>\n",
       "      <td>-0.147293</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>23</th>\n",
       "      <td>-0.115692</td>\n",
       "      <td>0.061643</td>\n",
       "      <td>0.277216</td>\n",
       "      <td>0.362107</td>\n",
       "      <td>-0.585175</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                    Social drinker  Distance from Residence to Work    Weight  \\\n",
       "Reason for absence                                                              \n",
       "26                        0.220163                         0.277262  0.023891   \n",
       "0                        -0.813139                         0.473745 -0.941796   \n",
       "23                        0.119235                        -0.130268 -0.052720   \n",
       "7                        -0.839607                         0.590692  0.207264   \n",
       "23                       -0.115692                         0.061643  0.277216   \n",
       "\n",
       "                    Education  Body mass index  labels  \n",
       "Reason for absence                                      \n",
       "26                   0.416488        -0.574165       2  \n",
       "0                    0.071819        -0.446411       6  \n",
       "23                   0.539152        -0.643213       2  \n",
       "7                    0.134280        -0.147293       0  \n",
       "23                   0.362107        -0.585175       0  "
      ]
     },
     "execution_count": 39,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Map the cluster labels back to each individual absenteeism and examine which were naturally grouped together\n",
    "absence_transformed['labels']=list(map(int, cluster_labels))\n",
    "absence_transformed.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "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>Social drinker</th>\n",
       "      <th>Distance from Residence to Work</th>\n",
       "      <th>Weight</th>\n",
       "      <th>Education</th>\n",
       "      <th>Body mass index</th>\n",
       "      <th>labels</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Reason for absence</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>22</th>\n",
       "      <td>0.168640</td>\n",
       "      <td>0.200160</td>\n",
       "      <td>-0.450106</td>\n",
       "      <td>0.484331</td>\n",
       "      <td>0.634288</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>23</th>\n",
       "      <td>-0.030780</td>\n",
       "      <td>-0.020388</td>\n",
       "      <td>-0.108434</td>\n",
       "      <td>0.353993</td>\n",
       "      <td>0.284006</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>10</th>\n",
       "      <td>-0.050468</td>\n",
       "      <td>0.298167</td>\n",
       "      <td>-0.049213</td>\n",
       "      <td>0.243537</td>\n",
       "      <td>0.644102</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>18</th>\n",
       "      <td>0.045163</td>\n",
       "      <td>0.521368</td>\n",
       "      <td>-0.352631</td>\n",
       "      <td>0.338169</td>\n",
       "      <td>0.385478</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>23</th>\n",
       "      <td>0.520496</td>\n",
       "      <td>0.189885</td>\n",
       "      <td>0.255469</td>\n",
       "      <td>0.403811</td>\n",
       "      <td>0.571548</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                    Social drinker  Distance from Residence to Work    Weight  \\\n",
       "Reason for absence                                                              \n",
       "22                        0.168640                         0.200160 -0.450106   \n",
       "23                       -0.030780                        -0.020388 -0.108434   \n",
       "10                       -0.050468                         0.298167 -0.049213   \n",
       "18                        0.045163                         0.521368 -0.352631   \n",
       "23                        0.520496                         0.189885  0.255469   \n",
       "\n",
       "                    Education  Body mass index  labels  \n",
       "Reason for absence                                      \n",
       "22                   0.484331         0.634288       1  \n",
       "23                   0.353993         0.284006       1  \n",
       "10                   0.243537         0.644102       1  \n",
       "18                   0.338169         0.385478       1  \n",
       "23                   0.403811         0.571548       1  "
      ]
     },
     "execution_count": 40,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Examine one of the clusters in more detail\n",
    "cluster=absence_transformed[absence_transformed['labels']==1]\n",
    "cluster.head(5)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Delete the PCA endpoint\n",
    "sagemaker.Session().delete_endpoint(pca_predictor.endpoint)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Delete the K-Means endpoint\n",
    "sagemaker.Session().delete_endpoint(kmeans_predictor.endpoint)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "conda_python3",
   "language": "python",
   "name": "conda_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": "3.6.5"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}