Breast Cancer Diagnosis.ipynb

{
    "nbformat": 4, 
    "cells": [
        {
            "cell_type": "markdown", 
            "metadata": {
                "collapsed": true
            }, 
            "source": "# Breast Cancer Diagnosis\n   \n   - Dataset, RDD APIs\n   - A dataframe is nothing but a table with columns, rows and headers\n   - In this notebook, we will work through a dataset to demostrate spark's SQL-like abilities. We will also look at its machine learning capabilities in context of this dataset."
        }, 
        {
            "cell_type": "markdown", 
            "metadata": {}, 
            "source": "# 1. Use case\n- The chosen dataset is that of Breast Cancer. This set is collected from digitized image of a fine needle aspirate (FNA) of a breast mass. For more information about this dataset, please visit https://www.kaggle.com/uciml/breast-cancer-wisconsin-data\n- Each row contains information about a single breast mass along with a diagnosis of malignant/benign for this mass\n    \n### Problem statement\n- We need to learn patterns describing malignant and benign masses and need to be able to place any future samples in either of these buckets.  \n- Thus, the problem is a classification one where given a set of probable results, we need to pick one result with confidence. \n    \n### Solution\n   We will go through the following steps:\n   - Read data (Spark SQL)\n   - Feature Engineering (Accumulators, Broadcasters, ml/mllib APIs)\n   - Data Visualizations (PixieDust, Seaborn)\n   - Modeling (Spark ML)\n   - Evaluation and prediction (Spark ML)\n   - Deployment (Watson ML repository)"
        }, 
        {
            "cell_type": "code", 
            "metadata": {}, 
            "source": "from IPython.display import display\nfrom IPython.core.display import HTML \n\n\n\nfrom IPython.display import HTML\n\nHTML('''<script>\ncode_show=true; \nfunction code_toggle() {\n if (code_show){\n $('div.input').hide();\n } else {\n $('div.input').show();\n }\n code_show = !code_show\n} \n$( document ).ready(code_toggle);\n</script>\n<form action=\"javascript:code_toggle()\"><input type=\"submit\" value=\"Click here to toggle on/off the raw code.\"></form>''')\n\n", 
            "outputs": [
                {
                    "metadata": {}, 
                    "data": {
                        "text/plain": "<IPython.core.display.HTML object>", 
                        "text/html": "<script>\ncode_show=true; \nfunction code_toggle() {\n if (code_show){\n $('div.input').hide();\n } else {\n $('div.input').show();\n }\n code_show = !code_show\n} \n$( document ).ready(code_toggle);\n</script>\n<form action=\"javascript:code_toggle()\"><input type=\"submit\" value=\"Click here to toggle on/off the raw code.\"></form>"
                    }, 
                    "execution_count": 6, 
                    "output_type": "execute_result"
                }
            ], 
            "execution_count": 6
        }, 
        {
            "cell_type": "code", 
            "metadata": {}, 
            "source": "!pip install seaborn", 
            "outputs": [
                {
                    "text": "Requirement already satisfied: seaborn in /gpfs/global_fs01/sym_shared/YPProdSpark/user/s280-a46e86a84e2274-95b1885704d4/.local/lib/python2.7/site-packages\r\n", 
                    "output_type": "stream", 
                    "name": "stdout"
                }
            ], 
            "execution_count": 69
        }, 
        {
            "cell_type": "markdown", 
            "metadata": {}, 
            "source": "#### Machine Learning process ###\n\n<center><img width=\"600px\" height=\"600px\" src=\"https://raw.githubusercontent.com/martinkearn/Content/master/Blogs/Images/MLProcess.PNG\"></center>"
        }, 
        {
            "cell_type": "markdown", 
            "metadata": {}, 
            "source": "### Get dataset\n!wget https://raw.githubusercontent.com/joshishwetha/dsx-spark/master/data.csv"
        }, 
        {
            "cell_type": "code", 
            "metadata": {}, 
            "source": "!wget https://raw.githubusercontent.com/joshishwetha/dsx-spark/master/data.csv", 
            "outputs": [
                {
                    "text": "--2017-06-23 15:48:51--  https://raw.githubusercontent.com/joshishwetha/dsx-spark/master/data.csv\nResolving raw.githubusercontent.com (raw.githubusercontent.com)... 151.101.48.133\nConnecting to raw.githubusercontent.com (raw.githubusercontent.com)|151.101.48.133|:443... connected.\nHTTP request sent, awaiting response... 200 OK\nLength: 125139 (122K) [text/plain]\nSaving to: \u2018data.csv.10\u2019\n\n100%[======================================>] 125,139     --.-K/s   in 0.004s  \n\n2017-06-23 15:48:51 (27.5 MB/s) - \u2018data.csv.10\u2019 saved [125139/125139]\n\n", 
                    "output_type": "stream", 
                    "name": "stdout"
                }
            ], 
            "execution_count": 70
        }, 
        {
            "cell_type": "markdown", 
            "metadata": {}, 
            "source": "# 2. Data Exploration"
        }, 
        {
            "cell_type": "code", 
            "metadata": {}, 
            "source": "data = spark.read.csv('data.csv',inferSchema='true',header='true')\ndata = data.drop('_c32')", 
            "outputs": [], 
            "execution_count": 5
        }, 
        {
            "cell_type": "markdown", 
            "metadata": {}, 
            "source": "<center><img width=\"450px\" height=\"450px\" margin=\"auto\" src=\"https://st2.depositphotos.com/1724125/10885/v/950/depositphotos_108850596-stock-illustration-cartoon-safari-kids.jpg\"></center>"
        }, 
        {
            "cell_type": "markdown", 
            "metadata": {}, 
            "source": "### What does our data look like?"
        }, 
        {
            "cell_type": "code", 
            "metadata": {}, 
            "source": "data.toPandas()[:10]", 
            "outputs": [
                {
                    "metadata": {}, 
                    "data": {
                        "text/plain": "         id diagnosis  radius_mean  texture_mean  perimeter_mean  area_mean  \\\n0    842302         M        17.99         10.38          122.80     1001.0   \n1    842517         M        20.57         17.77          132.90     1326.0   \n2  84300903         M        19.69         21.25          130.00     1203.0   \n3  84348301         M        11.42         20.38           77.58      386.1   \n4  84358402         M        20.29         14.34          135.10     1297.0   \n5    843786         M        12.45         15.70           82.57      477.1   \n6    844359         M        18.25         19.98          119.60     1040.0   \n7  84458202         M        13.71         20.83           90.20      577.9   \n8    844981         M        13.00         21.82           87.50      519.8   \n9  84501001         M        12.46         24.04           83.97      475.9   \n\n   smoothness_mean  compactness_mean  concavity_mean  concave points_mean  \\\n0          0.11840           0.27760         0.30010              0.14710   \n1          0.08474           0.07864         0.08690              0.07017   \n2          0.10960           0.15990         0.19740              0.12790   \n3          0.14250           0.28390         0.24140              0.10520   \n4          0.10030           0.13280         0.19800              0.10430   \n5          0.12780           0.17000         0.15780              0.08089   \n6          0.09463           0.10900         0.11270              0.07400   \n7          0.11890           0.16450         0.09366              0.05985   \n8          0.12730           0.19320         0.18590              0.09353   \n9          0.11860           0.23960         0.22730              0.08543   \n\n            ...             radius_worst  texture_worst  perimeter_worst  \\\n0           ...                    25.38          17.33           184.60   \n1           ...                    24.99          23.41           158.80   \n2           ...                    23.57          25.53           152.50   \n3           ...                    14.91          26.50            98.87   \n4           ...                    22.54          16.67           152.20   \n5           ...                    15.47          23.75           103.40   \n6           ...                    22.88          27.66           153.20   \n7           ...                    17.06          28.14           110.60   \n8           ...                    15.49          30.73           106.20   \n9           ...                    15.09          40.68            97.65   \n\n   area_worst  smoothness_worst  compactness_worst  concavity_worst  \\\n0      2019.0            0.1622             0.6656           0.7119   \n1      1956.0            0.1238             0.1866           0.2416   \n2      1709.0            0.1444             0.4245           0.4504   \n3       567.7            0.2098             0.8663           0.6869   \n4      1575.0            0.1374             0.2050           0.4000   \n5       741.6            0.1791             0.5249           0.5355   \n6      1606.0            0.1442             0.2576           0.3784   \n7       897.0            0.1654             0.3682           0.2678   \n8       739.3            0.1703             0.5401           0.5390   \n9       711.4            0.1853             1.0580           1.1050   \n\n   concave points_worst  symmetry_worst  fractal_dimension_worst  \n0                0.2654          0.4601                  0.11890  \n1                0.1860          0.2750                  0.08902  \n2                0.2430          0.3613                  0.08758  \n3                0.2575          0.6638                  0.17300  \n4                0.1625          0.2364                  0.07678  \n5                0.1741          0.3985                  0.12440  \n6                0.1932          0.3063                  0.08368  \n7                0.1556          0.3196                  0.11510  \n8                0.2060          0.4378                  0.10720  \n9                0.2210          0.4366                  0.20750  \n\n[10 rows x 32 columns]", 
                        "text/html": "<div>\n<style>\n    .dataframe thead tr:only-child th {\n        text-align: right;\n    }\n\n    .dataframe thead th {\n        text-align: left;\n    }\n\n    .dataframe tbody tr th {\n        vertical-align: top;\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>diagnosis</th>\n      <th>radius_mean</th>\n      <th>texture_mean</th>\n      <th>perimeter_mean</th>\n      <th>area_mean</th>\n      <th>smoothness_mean</th>\n      <th>compactness_mean</th>\n      <th>concavity_mean</th>\n      <th>concave points_mean</th>\n      <th>...</th>\n      <th>radius_worst</th>\n      <th>texture_worst</th>\n      <th>perimeter_worst</th>\n      <th>area_worst</th>\n      <th>smoothness_worst</th>\n      <th>compactness_worst</th>\n      <th>concavity_worst</th>\n      <th>concave points_worst</th>\n      <th>symmetry_worst</th>\n      <th>fractal_dimension_worst</th>\n    </tr>\n  </thead>\n  <tbody>\n    <tr>\n      <th>0</th>\n      <td>842302</td>\n      <td>M</td>\n      <td>17.99</td>\n      <td>10.38</td>\n      <td>122.80</td>\n      <td>1001.0</td>\n      <td>0.11840</td>\n      <td>0.27760</td>\n      <td>0.30010</td>\n      <td>0.14710</td>\n      <td>...</td>\n      <td>25.38</td>\n      <td>17.33</td>\n      <td>184.60</td>\n      <td>2019.0</td>\n      <td>0.1622</td>\n      <td>0.6656</td>\n      <td>0.7119</td>\n      <td>0.2654</td>\n      <td>0.4601</td>\n      <td>0.11890</td>\n    </tr>\n    <tr>\n      <th>1</th>\n      <td>842517</td>\n      <td>M</td>\n      <td>20.57</td>\n      <td>17.77</td>\n      <td>132.90</td>\n      <td>1326.0</td>\n      <td>0.08474</td>\n      <td>0.07864</td>\n      <td>0.08690</td>\n      <td>0.07017</td>\n      <td>...</td>\n      <td>24.99</td>\n      <td>23.41</td>\n      <td>158.80</td>\n      <td>1956.0</td>\n      <td>0.1238</td>\n      <td>0.1866</td>\n      <td>0.2416</td>\n      <td>0.1860</td>\n      <td>0.2750</td>\n      <td>0.08902</td>\n    </tr>\n    <tr>\n      <th>2</th>\n      <td>84300903</td>\n      <td>M</td>\n      <td>19.69</td>\n      <td>21.25</td>\n      <td>130.00</td>\n      <td>1203.0</td>\n      <td>0.10960</td>\n      <td>0.15990</td>\n      <td>0.19740</td>\n      <td>0.12790</td>\n      <td>...</td>\n      <td>23.57</td>\n      <td>25.53</td>\n      <td>152.50</td>\n      <td>1709.0</td>\n      <td>0.1444</td>\n      <td>0.4245</td>\n      <td>0.4504</td>\n      <td>0.2430</td>\n      <td>0.3613</td>\n      <td>0.08758</td>\n    </tr>\n    <tr>\n      <th>3</th>\n      <td>84348301</td>\n      <td>M</td>\n      <td>11.42</td>\n      <td>20.38</td>\n      <td>77.58</td>\n      <td>386.1</td>\n      <td>0.14250</td>\n      <td>0.28390</td>\n      <td>0.24140</td>\n      <td>0.10520</td>\n      <td>...</td>\n      <td>14.91</td>\n      <td>26.50</td>\n      <td>98.87</td>\n      <td>567.7</td>\n      <td>0.2098</td>\n      <td>0.8663</td>\n      <td>0.6869</td>\n      <td>0.2575</td>\n      <td>0.6638</td>\n      <td>0.17300</td>\n    </tr>\n    <tr>\n      <th>4</th>\n      <td>84358402</td>\n      <td>M</td>\n      <td>20.29</td>\n      <td>14.34</td>\n      <td>135.10</td>\n      <td>1297.0</td>\n      <td>0.10030</td>\n      <td>0.13280</td>\n      <td>0.19800</td>\n      <td>0.10430</td>\n      <td>...</td>\n      <td>22.54</td>\n      <td>16.67</td>\n      <td>152.20</td>\n      <td>1575.0</td>\n      <td>0.1374</td>\n      <td>0.2050</td>\n      <td>0.4000</td>\n      <td>0.1625</td>\n      <td>0.2364</td>\n      <td>0.07678</td>\n    </tr>\n    <tr>\n      <th>5</th>\n      <td>843786</td>\n      <td>M</td>\n      <td>12.45</td>\n      <td>15.70</td>\n      <td>82.57</td>\n      <td>477.1</td>\n      <td>0.12780</td>\n      <td>0.17000</td>\n      <td>0.15780</td>\n      <td>0.08089</td>\n      <td>...</td>\n      <td>15.47</td>\n      <td>23.75</td>\n      <td>103.40</td>\n      <td>741.6</td>\n      <td>0.1791</td>\n      <td>0.5249</td>\n      <td>0.5355</td>\n      <td>0.1741</td>\n      <td>0.3985</td>\n      <td>0.12440</td>\n    </tr>\n    <tr>\n      <th>6</th>\n      <td>844359</td>\n      <td>M</td>\n      <td>18.25</td>\n      <td>19.98</td>\n      <td>119.60</td>\n      <td>1040.0</td>\n      <td>0.09463</td>\n      <td>0.10900</td>\n      <td>0.11270</td>\n      <td>0.07400</td>\n      <td>...</td>\n      <td>22.88</td>\n      <td>27.66</td>\n      <td>153.20</td>\n      <td>1606.0</td>\n      <td>0.1442</td>\n      <td>0.2576</td>\n      <td>0.3784</td>\n      <td>0.1932</td>\n      <td>0.3063</td>\n      <td>0.08368</td>\n    </tr>\n    <tr>\n      <th>7</th>\n      <td>84458202</td>\n      <td>M</td>\n      <td>13.71</td>\n      <td>20.83</td>\n      <td>90.20</td>\n      <td>577.9</td>\n      <td>0.11890</td>\n      <td>0.16450</td>\n      <td>0.09366</td>\n      <td>0.05985</td>\n      <td>...</td>\n      <td>17.06</td>\n      <td>28.14</td>\n      <td>110.60</td>\n      <td>897.0</td>\n      <td>0.1654</td>\n      <td>0.3682</td>\n      <td>0.2678</td>\n      <td>0.1556</td>\n      <td>0.3196</td>\n      <td>0.11510</td>\n    </tr>\n    <tr>\n      <th>8</th>\n      <td>844981</td>\n      <td>M</td>\n      <td>13.00</td>\n      <td>21.82</td>\n      <td>87.50</td>\n      <td>519.8</td>\n      <td>0.12730</td>\n      <td>0.19320</td>\n      <td>0.18590</td>\n      <td>0.09353</td>\n      <td>...</td>\n      <td>15.49</td>\n      <td>30.73</td>\n      <td>106.20</td>\n      <td>739.3</td>\n      <td>0.1703</td>\n      <td>0.5401</td>\n      <td>0.5390</td>\n      <td>0.2060</td>\n      <td>0.4378</td>\n      <td>0.10720</td>\n    </tr>\n    <tr>\n      <th>9</th>\n      <td>84501001</td>\n      <td>M</td>\n      <td>12.46</td>\n      <td>24.04</td>\n      <td>83.97</td>\n      <td>475.9</td>\n      <td>0.11860</td>\n      <td>0.23960</td>\n      <td>0.22730</td>\n      <td>0.08543</td>\n      <td>...</td>\n      <td>15.09</td>\n      <td>40.68</td>\n      <td>97.65</td>\n      <td>711.4</td>\n      <td>0.1853</td>\n      <td>1.0580</td>\n      <td>1.1050</td>\n      <td>0.2210</td>\n      <td>0.4366</td>\n      <td>0.20750</td>\n    </tr>\n  </tbody>\n</table>\n<p>10 rows \u00d7 32 columns</p>\n</div>"
                    }, 
                    "execution_count": 7, 
                    "output_type": "execute_result"
                }
            ], 
            "execution_count": 7
        }, 
        {
            "cell_type": "markdown", 
            "metadata": {}, 
            "source": "### Check datatypes"
        }, 
        {
            "cell_type": "code", 
            "metadata": {}, 
            "source": "data.dtypes", 
            "outputs": [
                {
                    "metadata": {}, 
                    "data": {
                        "text/plain": "[('id', 'int'),\n ('diagnosis', 'string'),\n ('radius_mean', 'double'),\n ('texture_mean', 'double'),\n ('perimeter_mean', 'double'),\n ('area_mean', 'double'),\n ('smoothness_mean', 'double'),\n ('compactness_mean', 'double'),\n ('concavity_mean', 'double'),\n ('concave points_mean', 'double'),\n ('symmetry_mean', 'double'),\n ('fractal_dimension_mean', 'double'),\n ('radius_se', 'double'),\n ('texture_se', 'double'),\n ('perimeter_se', 'double'),\n ('area_se', 'double'),\n ('smoothness_se', 'double'),\n ('compactness_se', 'double'),\n ('concavity_se', 'double'),\n ('concave points_se', 'double'),\n ('symmetry_se', 'double'),\n ('fractal_dimension_se', 'double'),\n ('radius_worst', 'double'),\n ('texture_worst', 'double'),\n ('perimeter_worst', 'double'),\n ('area_worst', 'double'),\n ('smoothness_worst', 'double'),\n ('compactness_worst', 'double'),\n ('concavity_worst', 'double'),\n ('concave points_worst', 'double'),\n ('symmetry_worst', 'double'),\n ('fractal_dimension_worst', 'double')]"
                    }, 
                    "execution_count": 8, 
                    "output_type": "execute_result"
                }
            ], 
            "execution_count": 8
        }, 
        {
            "cell_type": "code", 
            "metadata": {
                "collapsed": true
            }, 
            "source": "from pyspark.sql.types import FloatType, StringType, IntegerType, DoubleType, ArrayType\n\nfor col in data.columns:\n    if col not in ['id','diagnosis']:\n        data = data.withColumn(col,data[col].cast(FloatType()))", 
            "outputs": [], 
            "execution_count": 9
        }, 
        {
            "cell_type": "code", 
            "metadata": {
                "collapsed": true
            }, 
            "source": "data = data.na.fill(0)", 
            "outputs": [], 
            "execution_count": 10
        }, 
        {
            "cell_type": "markdown", 
            "metadata": {}, 
            "source": "### SQL\n\n#### Register dataframe as a temp table to query from (write sql on dataframes)\n    data.registerTempTable('cancer_data')"
        }, 
        {
            "cell_type": "code", 
            "metadata": {
                "collapsed": true
            }, 
            "source": "data.registerTempTable('cancer_data')", 
            "outputs": [], 
            "execution_count": 11
        }, 
        {
            "cell_type": "markdown", 
            "metadata": {}, 
            "source": "#### Querying data\n- NOTE that the returned object is another Dataframe\n- One nice feature of the notebooks and python is that we can show it in a table via Pandas\n- Remember to perform an action to get your results (sql queries are also transformations :))"
        }, 
        {
            "cell_type": "markdown", 
            "metadata": {}, 
            "source": "    query = \"\"\"\n    select\n        diagnosis ,\n        count(1) as diagnosis_count\n    from cancer_data\n    group by diagnosis \n    \"\"\"\n    spark.sql(query).toPandas()\n"
        }, 
        {
            "cell_type": "code", 
            "metadata": {}, 
            "source": "query = \"\"\"\n    select\n        diagnosis ,\n        count(1) as diagnosis_count\n    from cancer_data\n    group by diagnosis \n    \"\"\"\nspark.sql(query).toPandas()", 
            "outputs": [
                {
                    "metadata": {}, 
                    "data": {
                        "text/plain": "  diagnosis  diagnosis_count\n0         B              357\n1         M              212", 
                        "text/html": "<div>\n<style>\n    .dataframe thead tr:only-child th {\n        text-align: right;\n    }\n\n    .dataframe thead th {\n        text-align: left;\n    }\n\n    .dataframe tbody tr th {\n        vertical-align: top;\n    }\n</style>\n<table border=\"1\" class=\"dataframe\">\n  <thead>\n    <tr style=\"text-align: right;\">\n      <th></th>\n      <th>diagnosis</th>\n      <th>diagnosis_count</th>\n    </tr>\n  </thead>\n  <tbody>\n    <tr>\n      <th>0</th>\n      <td>B</td>\n      <td>357</td>\n    </tr>\n    <tr>\n      <th>1</th>\n      <td>M</td>\n      <td>212</td>\n    </tr>\n  </tbody>\n</table>\n</div>"
                    }, 
                    "execution_count": 12, 
                    "output_type": "execute_result"
                }
            ], 
            "execution_count": 12
        }, 
        {
            "cell_type": "markdown", 
            "metadata": {}, 
            "source": "# 3. Feature Engineering (Data Preparation)\n   - Dataset = Features + target\n   - Transforming features (categorical to numeric, continous to bins, scaling, normalization etc...)\n   - Selecting a subset of columns for wide datasets\n   - Exploding columns to make additional (synthetic features) for small datasets\n   \n   \n<img width=\"450px\" height=\"450px\" margin=\"auto\" src=\"https://media.licdn.com/mpr/mpr/AAEAAQAAAAAAAAiuAAAAJGRkNGQzNjU5LTQzOTAtNDRlZC05NTZkLTcxZWM5YmYwZmRkNA.png\"></img>"
        }, 
        {
            "cell_type": "markdown", 
            "metadata": {}, 
            "source": "### Step 1: Make sure your target is numeric\n- \"target\" is categorical (malignant/benign)\n- Our encoding: 1 = malignant(M), 0 = Benign(B) using sql like \"when\" statement"
        }, 
        {
            "cell_type": "code", 
            "metadata": {
                "collapsed": true
            }, 
            "source": "#### Replace null values with 0\ndata = data.na.fill(0)", 
            "outputs": [], 
            "execution_count": 13
        }, 
        {
            "cell_type": "markdown", 
            "metadata": {}, 
            "source": "### Step 2: Sample your data to view what it looks like before we make changes to the target column\n\n    data.select('diagnosis','radius_se').show(3)"
        }, 
        {
            "cell_type": "code", 
            "metadata": {}, 
            "source": "data.select('diagnosis','radius_se').show(3)", 
            "outputs": [
                {
                    "text": "+---------+---------+\n|diagnosis|radius_se|\n+---------+---------+\n|        M|    1.095|\n|        M|   0.5435|\n|        M|   0.7456|\n+---------+---------+\nonly showing top 3 rows\n\n", 
                    "output_type": "stream", 
                    "name": "stdout"
                }
            ], 
            "execution_count": 14
        }, 
        {
            "cell_type": "markdown", 
            "metadata": {}, 
            "source": "#### Make label numeric"
        }, 
        {
            "cell_type": "code", 
            "metadata": {}, 
            "source": "import  pyspark.sql.functions as F\n\ndata = data.withColumn('diagnosis',F.when(data.diagnosis=='M',1).otherwise(0))\ndata.toPandas().head(3)", 
            "outputs": [
                {
                    "metadata": {}, 
                    "data": {
                        "text/plain": "         id  diagnosis  radius_mean  texture_mean  perimeter_mean  area_mean  \\\n0    842302          1    17.990000         10.38      122.800003     1001.0   \n1    842517          1    20.570000         17.77      132.899994     1326.0   \n2  84300903          1    19.690001         21.25      130.000000     1203.0   \n\n   smoothness_mean  compactness_mean  concavity_mean  concave points_mean  \\\n0          0.11840           0.27760          0.3001              0.14710   \n1          0.08474           0.07864          0.0869              0.07017   \n2          0.10960           0.15990          0.1974              0.12790   \n\n            ...             radius_worst  texture_worst  perimeter_worst  \\\n0           ...                25.379999      17.330000       184.600006   \n1           ...                24.990000      23.410000       158.800003   \n2           ...                23.570000      25.530001       152.500000   \n\n   area_worst  smoothness_worst  compactness_worst  concavity_worst  \\\n0      2019.0            0.1622             0.6656           0.7119   \n1      1956.0            0.1238             0.1866           0.2416   \n2      1709.0            0.1444             0.4245           0.4504   \n\n   concave points_worst  symmetry_worst  fractal_dimension_worst  \n0                0.2654          0.4601                  0.11890  \n1                0.1860          0.2750                  0.08902  \n2                0.2430          0.3613                  0.08758  \n\n[3 rows x 32 columns]", 
                        "text/html": "<div>\n<style>\n    .dataframe thead tr:only-child th {\n        text-align: right;\n    }\n\n    .dataframe thead th {\n        text-align: left;\n    }\n\n    .dataframe tbody tr th {\n        vertical-align: top;\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>diagnosis</th>\n      <th>radius_mean</th>\n      <th>texture_mean</th>\n      <th>perimeter_mean</th>\n      <th>area_mean</th>\n      <th>smoothness_mean</th>\n      <th>compactness_mean</th>\n      <th>concavity_mean</th>\n      <th>concave points_mean</th>\n      <th>...</th>\n      <th>radius_worst</th>\n      <th>texture_worst</th>\n      <th>perimeter_worst</th>\n      <th>area_worst</th>\n      <th>smoothness_worst</th>\n      <th>compactness_worst</th>\n      <th>concavity_worst</th>\n      <th>concave points_worst</th>\n      <th>symmetry_worst</th>\n      <th>fractal_dimension_worst</th>\n    </tr>\n  </thead>\n  <tbody>\n    <tr>\n      <th>0</th>\n      <td>842302</td>\n      <td>1</td>\n      <td>17.990000</td>\n      <td>10.38</td>\n      <td>122.800003</td>\n      <td>1001.0</td>\n      <td>0.11840</td>\n      <td>0.27760</td>\n      <td>0.3001</td>\n      <td>0.14710</td>\n      <td>...</td>\n      <td>25.379999</td>\n      <td>17.330000</td>\n      <td>184.600006</td>\n      <td>2019.0</td>\n      <td>0.1622</td>\n      <td>0.6656</td>\n      <td>0.7119</td>\n      <td>0.2654</td>\n      <td>0.4601</td>\n      <td>0.11890</td>\n    </tr>\n    <tr>\n      <th>1</th>\n      <td>842517</td>\n      <td>1</td>\n      <td>20.570000</td>\n      <td>17.77</td>\n      <td>132.899994</td>\n      <td>1326.0</td>\n      <td>0.08474</td>\n      <td>0.07864</td>\n      <td>0.0869</td>\n      <td>0.07017</td>\n      <td>...</td>\n      <td>24.990000</td>\n      <td>23.410000</td>\n      <td>158.800003</td>\n      <td>1956.0</td>\n      <td>0.1238</td>\n      <td>0.1866</td>\n      <td>0.2416</td>\n      <td>0.1860</td>\n      <td>0.2750</td>\n      <td>0.08902</td>\n    </tr>\n    <tr>\n      <th>2</th>\n      <td>84300903</td>\n      <td>1</td>\n      <td>19.690001</td>\n      <td>21.25</td>\n      <td>130.000000</td>\n      <td>1203.0</td>\n      <td>0.10960</td>\n      <td>0.15990</td>\n      <td>0.1974</td>\n      <td>0.12790</td>\n      <td>...</td>\n      <td>23.570000</td>\n      <td>25.530001</td>\n      <td>152.500000</td>\n      <td>1709.0</td>\n      <td>0.1444</td>\n      <td>0.4245</td>\n      <td>0.4504</td>\n      <td>0.2430</td>\n      <td>0.3613</td>\n      <td>0.08758</td>\n    </tr>\n  </tbody>\n</table>\n<p>3 rows \u00d7 32 columns</p>\n</div>"
                    }, 
                    "execution_count": 15, 
                    "output_type": "execute_result"
                }
            ], 
            "execution_count": 15
        }, 
        {
            "cell_type": "markdown", 
            "metadata": {}, 
            "source": "#### Rename column from \"diagnosis\" to \"label\""
        }, 
        {
            "cell_type": "code", 
            "metadata": {}, 
            "source": "data = data.withColumnRenamed('diagnosis','label')\ndata.toPandas().head(3)", 
            "outputs": [
                {
                    "metadata": {}, 
                    "data": {
                        "text/plain": "         id  label  radius_mean  texture_mean  perimeter_mean  area_mean  \\\n0    842302      1    17.990000         10.38      122.800003     1001.0   \n1    842517      1    20.570000         17.77      132.899994     1326.0   \n2  84300903      1    19.690001         21.25      130.000000     1203.0   \n\n   smoothness_mean  compactness_mean  concavity_mean  concave points_mean  \\\n0          0.11840           0.27760          0.3001              0.14710   \n1          0.08474           0.07864          0.0869              0.07017   \n2          0.10960           0.15990          0.1974              0.12790   \n\n            ...             radius_worst  texture_worst  perimeter_worst  \\\n0           ...                25.379999      17.330000       184.600006   \n1           ...                24.990000      23.410000       158.800003   \n2           ...                23.570000      25.530001       152.500000   \n\n   area_worst  smoothness_worst  compactness_worst  concavity_worst  \\\n0      2019.0            0.1622             0.6656           0.7119   \n1      1956.0            0.1238             0.1866           0.2416   \n2      1709.0            0.1444             0.4245           0.4504   \n\n   concave points_worst  symmetry_worst  fractal_dimension_worst  \n0                0.2654          0.4601                  0.11890  \n1                0.1860          0.2750                  0.08902  \n2                0.2430          0.3613                  0.08758  \n\n[3 rows x 32 columns]", 
                        "text/html": "<div>\n<style>\n    .dataframe thead tr:only-child th {\n        text-align: right;\n    }\n\n    .dataframe thead th {\n        text-align: left;\n    }\n\n    .dataframe tbody tr th {\n        vertical-align: top;\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>label</th>\n      <th>radius_mean</th>\n      <th>texture_mean</th>\n      <th>perimeter_mean</th>\n      <th>area_mean</th>\n      <th>smoothness_mean</th>\n      <th>compactness_mean</th>\n      <th>concavity_mean</th>\n      <th>concave points_mean</th>\n      <th>...</th>\n      <th>radius_worst</th>\n      <th>texture_worst</th>\n      <th>perimeter_worst</th>\n      <th>area_worst</th>\n      <th>smoothness_worst</th>\n      <th>compactness_worst</th>\n      <th>concavity_worst</th>\n      <th>concave points_worst</th>\n      <th>symmetry_worst</th>\n      <th>fractal_dimension_worst</th>\n    </tr>\n  </thead>\n  <tbody>\n    <tr>\n      <th>0</th>\n      <td>842302</td>\n      <td>1</td>\n      <td>17.990000</td>\n      <td>10.38</td>\n      <td>122.800003</td>\n      <td>1001.0</td>\n      <td>0.11840</td>\n      <td>0.27760</td>\n      <td>0.3001</td>\n      <td>0.14710</td>\n      <td>...</td>\n      <td>25.379999</td>\n      <td>17.330000</td>\n      <td>184.600006</td>\n      <td>2019.0</td>\n      <td>0.1622</td>\n      <td>0.6656</td>\n      <td>0.7119</td>\n      <td>0.2654</td>\n      <td>0.4601</td>\n      <td>0.11890</td>\n    </tr>\n    <tr>\n      <th>1</th>\n      <td>842517</td>\n      <td>1</td>\n      <td>20.570000</td>\n      <td>17.77</td>\n      <td>132.899994</td>\n      <td>1326.0</td>\n      <td>0.08474</td>\n      <td>0.07864</td>\n      <td>0.0869</td>\n      <td>0.07017</td>\n      <td>...</td>\n      <td>24.990000</td>\n      <td>23.410000</td>\n      <td>158.800003</td>\n      <td>1956.0</td>\n      <td>0.1238</td>\n      <td>0.1866</td>\n      <td>0.2416</td>\n      <td>0.1860</td>\n      <td>0.2750</td>\n      <td>0.08902</td>\n    </tr>\n    <tr>\n      <th>2</th>\n      <td>84300903</td>\n      <td>1</td>\n      <td>19.690001</td>\n      <td>21.25</td>\n      <td>130.000000</td>\n      <td>1203.0</td>\n      <td>0.10960</td>\n      <td>0.15990</td>\n      <td>0.1974</td>\n      <td>0.12790</td>\n      <td>...</td>\n      <td>23.570000</td>\n      <td>25.530001</td>\n      <td>152.500000</td>\n      <td>1709.0</td>\n      <td>0.1444</td>\n      <td>0.4245</td>\n      <td>0.4504</td>\n      <td>0.2430</td>\n      <td>0.3613</td>\n      <td>0.08758</td>\n    </tr>\n  </tbody>\n</table>\n<p>3 rows \u00d7 32 columns</p>\n</div>"
                    }, 
                    "execution_count": 16, 
                    "output_type": "execute_result"
                }
            ], 
            "execution_count": 16
        }, 
        {
            "cell_type": "code", 
            "metadata": {
                "collapsed": true
            }, 
            "source": "def count_labels(row):\n    global benign, malignant\n    if row.label==1:\n        malignant.add(1)\n    else:\n        benign.add(1)", 
            "outputs": [], 
            "execution_count": 82
        }, 
        {
            "cell_type": "markdown", 
            "metadata": {}, 
            "source": "#### Collect statistics for your dataset ###\n- Collect statistics for mean/std\n- store it in a variable called \"df\""
        }, 
        {
            "cell_type": "code", 
            "metadata": {
                "scrolled": true
            }, 
            "source": "df = data.describe()\ndf.toPandas()", 
            "outputs": [
                {
                    "metadata": {}, 
                    "data": {
                        "text/plain": "  summary                    id                label         radius_mean  \\\n0   count                   569                  569                 569   \n1    mean  3.0371831432337433E7  0.37258347978910367  14.127291743072348   \n2  stddev  1.2502058561222367E8   0.4839179564031687   3.524048812967195   \n3     min                  8670                    0               6.981   \n4     max             911320502                    1               28.11   \n\n         texture_mean      perimeter_mean           area_mean  \\\n0                 569                 569                 569   \n1  19.289648528677297   91.96903329993384    654.889103814043   \n2   4.301035792275387  24.298980946187076  351.91412886139733   \n3                9.71               43.79               143.5   \n4               39.28               188.5              2501.0   \n\n        smoothness_mean     compactness_mean       concavity_mean  \\\n0                   569                  569                  569   \n1   0.09636028129312821  0.10434098429781481  0.08879931578830029   \n2  0.014064128011679866  0.05281275807458227  0.07971980885275734   \n3               0.05263              0.01938                  0.0   \n4                0.1634               0.3454               0.4268   \n\n            ...                  radius_worst      texture_worst  \\\n0           ...                           569                569   \n1           ...            16.269189776770887  25.67722316534113   \n2           ...            4.8332415912724365  6.146257611231105   \n3           ...                          7.93              12.02   \n4           ...                         36.04              49.54   \n\n      perimeter_worst         area_worst      smoothness_worst  \\\n0                 569                569                   569   \n1  107.26121279644421  880.5831290514901   0.13236859435565862   \n2  33.602542264508905  569.3569923849643  0.022832429559187094   \n3               50.41              185.2               0.07117   \n4               251.2             4254.0                0.2226   \n\n     compactness_worst      concavity_worst concave points_worst  \\\n0                  569                  569                  569   \n1  0.25426504394016597   0.2721884833807977  0.11460622294146325   \n2  0.15733648854662943  0.20862428007810724  0.06573234105890066   \n3              0.02729                  0.0                  0.0   \n4                1.058                1.252                0.291   \n\n        symmetry_worst fractal_dimension_worst  \n0                  569                     569  \n1   0.2900755708948799     0.08394581713895387  \n2  0.06186746818484165     0.01806126727675421  \n3               0.1565                 0.05504  \n4               0.6638                  0.2075  \n\n[5 rows x 33 columns]", 
                        "text/html": "<div>\n<style>\n    .dataframe thead tr:only-child th {\n        text-align: right;\n    }\n\n    .dataframe thead th {\n        text-align: left;\n    }\n\n    .dataframe tbody tr th {\n        vertical-align: top;\n    }\n</style>\n<table border=\"1\" class=\"dataframe\">\n  <thead>\n    <tr style=\"text-align: right;\">\n      <th></th>\n      <th>summary</th>\n      <th>id</th>\n      <th>label</th>\n      <th>radius_mean</th>\n      <th>texture_mean</th>\n      <th>perimeter_mean</th>\n      <th>area_mean</th>\n      <th>smoothness_mean</th>\n      <th>compactness_mean</th>\n      <th>concavity_mean</th>\n      <th>...</th>\n      <th>radius_worst</th>\n      <th>texture_worst</th>\n      <th>perimeter_worst</th>\n      <th>area_worst</th>\n      <th>smoothness_worst</th>\n      <th>compactness_worst</th>\n      <th>concavity_worst</th>\n      <th>concave points_worst</th>\n      <th>symmetry_worst</th>\n      <th>fractal_dimension_worst</th>\n    </tr>\n  </thead>\n  <tbody>\n    <tr>\n      <th>0</th>\n      <td>count</td>\n      <td>569</td>\n      <td>569</td>\n      <td>569</td>\n      <td>569</td>\n      <td>569</td>\n      <td>569</td>\n      <td>569</td>\n      <td>569</td>\n      <td>569</td>\n      <td>...</td>\n      <td>569</td>\n      <td>569</td>\n      <td>569</td>\n      <td>569</td>\n      <td>569</td>\n      <td>569</td>\n      <td>569</td>\n      <td>569</td>\n      <td>569</td>\n      <td>569</td>\n    </tr>\n    <tr>\n      <th>1</th>\n      <td>mean</td>\n      <td>3.0371831432337433E7</td>\n      <td>0.37258347978910367</td>\n      <td>14.127291743072348</td>\n      <td>19.289648528677297</td>\n      <td>91.96903329993384</td>\n      <td>654.889103814043</td>\n      <td>0.09636028129312821</td>\n      <td>0.10434098429781481</td>\n      <td>0.08879931578830029</td>\n      <td>...</td>\n      <td>16.269189776770887</td>\n      <td>25.67722316534113</td>\n      <td>107.26121279644421</td>\n      <td>880.5831290514901</td>\n      <td>0.13236859435565862</td>\n      <td>0.25426504394016597</td>\n      <td>0.2721884833807977</td>\n      <td>0.11460622294146325</td>\n      <td>0.2900755708948799</td>\n      <td>0.08394581713895387</td>\n    </tr>\n    <tr>\n      <th>2</th>\n      <td>stddev</td>\n      <td>1.2502058561222367E8</td>\n      <td>0.4839179564031687</td>\n      <td>3.524048812967195</td>\n      <td>4.301035792275387</td>\n      <td>24.298980946187076</td>\n      <td>351.91412886139733</td>\n      <td>0.014064128011679866</td>\n      <td>0.05281275807458227</td>\n      <td>0.07971980885275734</td>\n      <td>...</td>\n      <td>4.8332415912724365</td>\n      <td>6.146257611231105</td>\n      <td>33.602542264508905</td>\n      <td>569.3569923849643</td>\n      <td>0.022832429559187094</td>\n      <td>0.15733648854662943</td>\n      <td>0.20862428007810724</td>\n      <td>0.06573234105890066</td>\n      <td>0.06186746818484165</td>\n      <td>0.01806126727675421</td>\n    </tr>\n    <tr>\n      <th>3</th>\n      <td>min</td>\n      <td>8670</td>\n      <td>0</td>\n      <td>6.981</td>\n      <td>9.71</td>\n      <td>43.79</td>\n      <td>143.5</td>\n      <td>0.05263</td>\n      <td>0.01938</td>\n      <td>0.0</td>\n      <td>...</td>\n      <td>7.93</td>\n      <td>12.02</td>\n      <td>50.41</td>\n      <td>185.2</td>\n      <td>0.07117</td>\n      <td>0.02729</td>\n      <td>0.0</td>\n      <td>0.0</td>\n      <td>0.1565</td>\n      <td>0.05504</td>\n    </tr>\n    <tr>\n      <th>4</th>\n      <td>max</td>\n      <td>911320502</td>\n      <td>1</td>\n      <td>28.11</td>\n      <td>39.28</td>\n      <td>188.5</td>\n      <td>2501.0</td>\n      <td>0.1634</td>\n      <td>0.3454</td>\n      <td>0.4268</td>\n      <td>...</td>\n      <td>36.04</td>\n      <td>49.54</td>\n      <td>251.2</td>\n      <td>4254.0</td>\n      <td>0.2226</td>\n      <td>1.058</td>\n      <td>1.252</td>\n      <td>0.291</td>\n      <td>0.6638</td>\n      <td>0.2075</td>\n    </tr>\n  </tbody>\n</table>\n<p>5 rows \u00d7 33 columns</p>\n</div>"
                    }, 
                    "execution_count": 17, 
                    "output_type": "execute_result"
                }
            ], 
            "execution_count": 17
        }, 
        {
            "cell_type": "markdown", 
            "metadata": {}, 
            "source": "#### Calculate correlations of each columns w.r.t the target\n- Change to do this in spark\n\n\n\n#### Select a subset of columns\n- Remove columns conveying very little information\n   - Zero variance columns\n   - Low correlation columns"
        }, 
        {
            "cell_type": "code", 
            "metadata": {
                "collapsed": true
            }, 
            "source": "df = data.toPandas()\ncol_corr = sorted(df.corr()['label'].to_dict().items(),key=lambda x:x[1],reverse=True)\nfor col in col_corr[-5:]:\n    if col[0]!='id':\n        data = data.drop(col[0])", 
            "outputs": [], 
            "execution_count": 44
        }, 
        {
            "cell_type": "markdown", 
            "metadata": {}, 
            "source": "- Remove columns with low correlation, such columns contribute to very little information\n- Caution: Be wary of multicolinearity (\"leaky\" columns)"
        }, 
        {
            "cell_type": "markdown", 
            "metadata": {}, 
            "source": "## Vizualise Data (Lets do some graphs!)\n\n###  Here, we look at the columns and their correlations with the target variable.\n- We use PixieDust (charting library) for quick charts\n- Also takes RDDs as inputs, most of the other charting librabries take pandas dataframes as inputs"
        }, 
        {
            "cell_type": "code", 
            "metadata": {
                "scrolled": true, 
                "pixiedust": {
                    "displayParams": {
                        "legend": "false", 
                        "keyFields": "name", 
                        "handlerId": "barChart", 
                        "rowCount": "100", 
                        "valueFields": "correlation", 
                        "aggregation": "SUM"
                    }
                }
            }, 
            "source": "import pixiedust\nimport pandas as pd\n\ndf = pd.DataFrame(col_corr)\ndf = df.dropna()\ndf.columns = ['name','correlation']\nviz_spark_df = sqlContext.createDataFrame(df)\n\n\ndisplay(viz_spark_df)", 
            "outputs": [
                {
                    "text": "Pixiedust database opened successfully\n", 
                    "output_type": "stream", 
                    "name": "stdout"
                }, 
                {
                    "metadata": {}, 
                    "data": {
                        "text/plain": "<IPython.core.display.HTML object>", 
                        "text/html": "\n        <div style=\"margin:10px\">\n            <a href=\"https://github.com/ibm-cds-labs/pixiedust\" target=\"_new\">\n                <img src=\"https://github.com/ibm-cds-labs/pixiedust/raw/master/docs/_static/pd_icon32.png\" style=\"float:left;margin-right:10px\"/>\n            </a>\n            <span>Pixiedust version upgraded from 1.0.4 to 1.0.7</span>\n        </div>\n        "
                    }, 
                    "output_type": "display_data"
                }, 
                {
                    "traceback": [
                        "\u001b[0;31m\u001b[0m", 
                        "\u001b[0;31mNameError\u001b[0mTraceback (most recent call last)", 
                        "\u001b[0;32m<ipython-input-2-b24b8477012e>\u001b[0m in \u001b[0;36m<module>\u001b[0;34m()\u001b[0m\n\u001b[1;32m      2\u001b[0m \u001b[0;32mimport\u001b[0m \u001b[0mpandas\u001b[0m \u001b[0;32mas\u001b[0m \u001b[0mpd\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m      3\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m----> 4\u001b[0;31m \u001b[0mdf\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mpd\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mDataFrame\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mcol_corr\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m      5\u001b[0m \u001b[0mdf\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mdf\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mdropna\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m      6\u001b[0m \u001b[0mdf\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mcolumns\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0;34m[\u001b[0m\u001b[0;34m'name'\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m'correlation'\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n", 
                        "\u001b[0;31mNameError\u001b[0m: name 'col_corr' is not defined"
                    ], 
                    "output_type": "error", 
                    "ename": "NameError", 
                    "evalue": "name 'col_corr' is not defined"
                }
            ], 
            "execution_count": 2
        }, 
        {
            "cell_type": "markdown", 
            "metadata": {}, 
            "source": "### Visualize correlations ####\n - Useful for datasets with small number of columns\n - A heatmap is useful to visualize how variables are related to each other (and not just the target)\n - Very useful while doing NLP applications to visualize similarity between documents (after using TFIDF)\n "
        }, 
        {
            "cell_type": "code", 
            "metadata": {
                "collapsed": true
            }, 
            "source": "import matplotlib.pyplot as plt\nimport seaborn as sns\n\nfrom pyspark.mllib.stat import Statistics\n\nplt.style.use('ggplot')\nvalues = data.rdd.map(lambda x:list(x.asDict().values()))\ncorr_values = Statistics.corr(values)\nnames = data.rdd.map(lambda x:list(x.asDict().keys())).first()\n\nplt.figure(figsize=(10,10))\nsns.set_style(\"darkgrid\")\nsns.heatmap(corr_values,xticklabels=names,yticklabels=names,square=True,vmin=0, vmax=1,\n                cmap=\"YlGnBu\")", 
            "outputs": [], 
            "execution_count": 21
        }, 
        {
            "cell_type": "code", 
            "metadata": {
                "collapsed": true
            }, 
            "source": "###### PCA Visualize data ######\nfrom pyspark.ml.feature import PCA\nfrom pyspark.ml.linalg import Vectors\nfrom pyspark.sql.functions import UserDefinedFunction\nfrom pyspark.ml.feature import VectorAssembler \n\nfeature_cols = list(filter(lambda x:x not in ['id','label','dummy'],data.columns))\nassembler = VectorAssembler(inputCols=feature_cols,outputCol='features')\ndf = assembler.transform(data)\n\npca = PCA(k=2, inputCol=\"features\", outputCol=\"pcaFeatures\")\nmodel = pca.fit(df)\n\npca_result = model.transform(df)\nresult = pca_result.select(\"pcaFeatures\")\n", 
            "outputs": [], 
            "execution_count": 20
        }, 
        {
            "cell_type": "markdown", 
            "metadata": {}, 
            "source": "# 4. We are all set to do some machine learning!\n\n\n<center><img width=\"350px\" height=\"350px\" margin=\"auto\" src=\"http://www.clipartkid.com/images/127/cartoon-explorer-characters-vectors-6cbSmI-clipart.jpg\"></center>"
        }, 
        {
            "cell_type": "markdown", 
            "metadata": {}, 
            "source": "## Vizualise problem\n### PCA  - Dimensionality Reduction "
        }, 
        {
            "cell_type": "code", 
            "metadata": {
                "collapsed": true
            }, 
            "source": "%matplotlib inline", 
            "outputs": [], 
            "execution_count": 23
        }, 
        {
            "cell_type": "code", 
            "metadata": {}, 
            "source": "import regex\n\ndef g(point,co_ord):\n    if co_ord=='x':\n        value = regex.split('\\s+',point)[0][1:]\n    else:\n        value =  regex.split('\\s+',point)[1][:-1]\n    return value\n\nget_x_coord = UserDefinedFunction(lambda point:g(point,'x'), StringType())\nget_y_coord = UserDefinedFunction(lambda point:g(point,'y'), StringType())\nconvert_to_string = UserDefinedFunction(lambda point:str(point.toArray()),StringType())\n\npca_result = pca_result.withColumn('pcaFeatures_string',convert_to_string(pca_result['pcaFeatures']))\npca_result = pca_result.withColumn('x_coord',get_x_coord(pca_result['pcaFeatures_string']))\\\n                       .withColumn('y_coord',get_y_coord(pca_result['pcaFeatures_string']))\\\n                      \npca_result = pca_result.withColumn('x_coord',pca_result.x_coord.cast(FloatType()))\\\n                       .withColumn('y_coord',pca_result.y_coord.cast(FloatType()))\n    \nviz_df = pca_result.select('x_coord','y_coord','label').toPandas()\n\n\na = sns.lmplot(x='x_coord',y='y_coord',hue='label',data=viz_df,\n           fit_reg=False,palette={1:'red',0:'green'},size=6,aspect=1)", 
            "outputs": [
                {
                    "metadata": {}, 
                    "data": {
                        "text/plain": "<matplotlib.figure.Figure at 0x7fafec900b10>", 
                        "image/png": "iVBORw0KGgoAAAANSUhEUgAAAdIAAAGoCAYAAADy2WSRAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nH5+a+K+1EecnAAAAAElFTkSuQmCC\n"
                    }, 
                    "output_type": "display_data"
                }
            ], 
            "execution_count": 24
        }, 
        {
            "cell_type": "code", 
            "metadata": {
                "collapsed": true
            }, 
            "source": "#### Convert 'label' to DoubleType()\ndata = data.withColumn('label',data['label'].cast(DoubleType()))", 
            "outputs": [], 
            "execution_count": 25
        }, 
        {
            "cell_type": "code", 
            "metadata": {
                "collapsed": true
            }, 
            "source": "#### Split dataset into test and train sets (70/30) splits\ntrain, test = data.randomSplit([0.7,0.3])", 
            "outputs": [], 
            "execution_count": 26
        }, 
        {
            "cell_type": "code", 
            "metadata": {}, 
            "source": "#### Count number of records in test and train sets\ntrain.count(), test.count()", 
            "outputs": [
                {
                    "metadata": {}, 
                    "data": {
                        "text/plain": "(400, 169)"
                    }, 
                    "execution_count": 27, 
                    "output_type": "execute_result"
                }
            ], 
            "execution_count": 27
        }, 
        {
            "cell_type": "code", 
            "metadata": {
                "collapsed": true
            }, 
            "source": "from pyspark.ml.pipeline import Pipeline", 
            "outputs": [], 
            "execution_count": 28
        }, 
        {
            "cell_type": "code", 
            "metadata": {
                "collapsed": true
            }, 
            "source": "#### Define Assembler to combine feature into a single vetor (feature vector! :) )\nfrom pyspark.ml.linalg import Vectors #linear algebra package, has matrices, arrays, Vectors (dense and sparse)\nfrom pyspark.ml.feature import VectorAssembler \n\nfeature_cols = list(filter(lambda x:x not in ['id','label'],data.columns))\nassembler = VectorAssembler(inputCols=feature_cols,outputCol='features')", 
            "outputs": [], 
            "execution_count": 29
        }, 
        {
            "cell_type": "markdown", 
            "metadata": {}, 
            "source": "### Classification Pipelines \n- Tree based classifier\n- Logistic Regression\n\n<center><img src=\"https://codesachin.files.wordpress.com/2015/08/linearly_separable_4.png\"></center>"
        }, 
        {
            "cell_type": "markdown", 
            "metadata": {}, 
            "source": "### Results from a decision tree classifier"
        }, 
        {
            "cell_type": "code", 
            "metadata": {}, 
            "source": "from pyspark.ml.classification import DecisionTreeClassifier\n# define estimator and fit data\nestimator = DecisionTreeClassifier()\npipeline = Pipeline(stages=[assembler,estimator])\ntree_model = pipeline.fit(train)\n\n# get results\nresults = tree_model.transform(test)\ndt_classider = results\n\n### Visualize the results of tree classifier ###\npca_subset = pca_result.select('ID','features','x_coord','y_coord')\nviz_df = pca_subset.join(results,on=['ID','features']).select('x_coord','y_coord','prediction','label')\nviz_df = viz_df.dropna(how='any',subset=['x_coord','y_coord']).toPandas()\n\na = sns.lmplot(x='x_coord',y='y_coord',hue='prediction',col='label',data=viz_df,\n           fit_reg=False,palette={1:'red',0:'green'},size=6,aspect=1)", 
            "outputs": [
                {
                    "metadata": {}, 
                    "data": {
                        "text/plain": "<matplotlib.figure.Figure at 0x7fafec34a410>", 
                        "image/png": "iVBORw0KGgoAAAANSUhEUgAAA54AAAGoCAYAAAAuMzp+AAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n"
                    }, 
                    "output_type": "display_data"
                }
            ], 
            "execution_count": 30
        }, 
        {
            "cell_type": "code", 
            "metadata": {
                "collapsed": true
            }, 
            "source": "#### Classification metrics ####\ndef calc_metrics(results):\n    metrics = {}\n    metrics['tp_0'] = results.filter((results.label==0)&(results.prediction==0)).count()\n    metrics['fn_0'] = results.filter((results.label==0)&(results.prediction==1)).count()\n    metrics['tn_0'] = results.filter((results.label==1)&(results.prediction==1)).count()\n    metrics['fp_0'] = results.filter((results.label==1)&(results.prediction==0)).count()\n    \n    metrics['tp_1'] = results.filter((results.label==1)&(results.prediction==1)).count()\n    metrics['fn_1'] = results.filter((results.label==1)&(results.prediction==0)).count()\n    metrics['tn_1'] = results.filter((results.label==0)&(results.prediction==0)).count()\n    metrics['fp_1'] = results.filter((results.label==0)&(results.prediction==1)).count()\n    \n    return metrics\n        \n### calc precision & recall ###\ndef precision_recall(results):\n    items = calc_metrics(results)\n    rows = []\n    \n    row = {'class':0,'precision':items['tp_0']/float((items['tp_0']+items['fp_0'])+1),'recall':items['tp_0']/float((items['tp_0']+items['fn_0'])+1)}\n    rows.append(row)\n    row = {'class':1,'precision':items['tp_1']/float((items['tp_1']+items['fp_1'])+1),'recall':items['tp_1']/float((items['tp_1']+items['fn_1'])+1)}\n    rows.append(row)\n    \n    return rows", 
            "outputs": [], 
            "execution_count": 37
        }, 
        {
            "cell_type": "code", 
            "metadata": {
                "collapsed": true
            }, 
            "source": "from pyspark.ml.classification import LogisticRegression\n\n#define estimator and fit data\nestimator = LogisticRegression()\npipeline = Pipeline(stages=[assembler,estimator])\nlr_model = pipeline.fit(train)\n\n#get results\nresults = lr_model.transform(test)", 
            "outputs": [], 
            "execution_count": 32
        }, 
        {
            "cell_type": "code", 
            "metadata": {
                "collapsed": true
            }, 
            "source": "lr_results = results", 
            "outputs": [], 
            "execution_count": 33
        }, 
        {
            "cell_type": "markdown", 
            "metadata": {}, 
            "source": "### Results from Logistic regression classifier"
        }, 
        {
            "cell_type": "code", 
            "metadata": {}, 
            "source": "### Visualize results from logistic regression ###\npca_subset = pca_result.select('id','features','x_coord','y_coord')\nviz_df = pca_subset.join(results,on=['id','features']).select('x_coord','y_coord','prediction','label')\nviz_df = viz_df.dropna(how='any',subset=['x_coord','y_coord']).toPandas()\n\na = sns.lmplot(x='x_coord',y='y_coord',hue='prediction',col='label',data=viz_df,\n           fit_reg=False,palette={1:'red',0:'green'},size=6,aspect=1)\na.set_titles(col_template=['Decision Tree results when label==\"Benign\"','Decision Tree results when label==\"Malignant\"'])", 
            "outputs": [
                {
                    "metadata": {}, 
                    "data": {
                        "text/plain": "<seaborn.axisgrid.FacetGrid at 0x7fafec083dd0>"
                    }, 
                    "execution_count": 34, 
                    "output_type": "execute_result"
                }, 
                {
                    "metadata": {}, 
                    "data": {
                        "text/plain": "<matplotlib.figure.Figure at 0x7fafec8e6fd0>", 
                        "image/png": "iVBORw0KGgoAAAANSUhEUgAAA+cAAAGoCAYAAAAkb5QsAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n/wAAAABJRU5ErkJggg==\n"
                    }, 
                    "output_type": "display_data"
                }
            ], 
            "execution_count": 34
        }, 
        {
            "cell_type": "markdown", 
            "metadata": {}, 
            "source": "### Evaluation\n\n- Classification models work towards making a Yes/No decision\n- Metrics to evaluate the model: Precision/Recall, AUC\n\n\n<center><img width=\"500px\" height=\"500px\" src=\"http://opexanalytics.com/cnt/uploads/2016/01/Red-Fish-High-Recall.jpg\"></center>"
        }, 
        {
            "cell_type": "code", 
            "metadata": {}, 
            "source": "### Precision Recall scores from Decision Tree Classifier\nimport pandas as pd\npd.DataFrame(precision_recall(dt_classider))", 
            "outputs": [
                {
                    "metadata": {}, 
                    "data": {
                        "text/plain": "   class  precision    recall\n0      0   0.915888  0.915888\n1      1   0.859375  0.859375", 
                        "text/html": "<div>\n<style>\n    .dataframe thead tr:only-child th {\n        text-align: right;\n    }\n\n    .dataframe thead th {\n        text-align: left;\n    }\n\n    .dataframe tbody tr th {\n        vertical-align: top;\n    }\n</style>\n<table border=\"1\" class=\"dataframe\">\n  <thead>\n    <tr style=\"text-align: right;\">\n      <th></th>\n      <th>class</th>\n      <th>precision</th>\n      <th>recall</th>\n    </tr>\n  </thead>\n  <tbody>\n    <tr>\n      <th>0</th>\n      <td>0</td>\n      <td>0.915888</td>\n      <td>0.915888</td>\n    </tr>\n    <tr>\n      <th>1</th>\n      <td>1</td>\n      <td>0.859375</td>\n      <td>0.859375</td>\n    </tr>\n  </tbody>\n</table>\n</div>"
                    }, 
                    "execution_count": 41, 
                    "output_type": "execute_result"
                }
            ], 
            "execution_count": 41
        }, 
        {
            "cell_type": "markdown", 
            "metadata": {}, 
            "source": "### Precision Recall scores from Logistic Regression Classifier"
        }, 
        {
            "cell_type": "code", 
            "metadata": {}, 
            "source": "pd.DataFrame(precision_recall(lr_results))", 
            "outputs": [
                {
                    "metadata": {}, 
                    "data": {
                        "text/plain": "   class  precision    recall\n0      0   0.926606  0.943925\n1      1   0.903226  0.875000", 
                        "text/html": "<div>\n<style>\n    .dataframe thead tr:only-child th {\n        text-align: right;\n    }\n\n    .dataframe thead th {\n        text-align: left;\n    }\n\n    .dataframe tbody tr th {\n        vertical-align: top;\n    }\n</style>\n<table border=\"1\" class=\"dataframe\">\n  <thead>\n    <tr style=\"text-align: right;\">\n      <th></th>\n      <th>class</th>\n      <th>precision</th>\n      <th>recall</th>\n    </tr>\n  </thead>\n  <tbody>\n    <tr>\n      <th>0</th>\n      <td>0</td>\n      <td>0.926606</td>\n      <td>0.943925</td>\n    </tr>\n    <tr>\n      <th>1</th>\n      <td>1</td>\n      <td>0.903226</td>\n      <td>0.875000</td>\n    </tr>\n  </tbody>\n</table>\n</div>"
                    }, 
                    "execution_count": 42, 
                    "output_type": "execute_result"
                }
            ], 
            "execution_count": 42
        }, 
        {
            "cell_type": "markdown", 
            "metadata": {}, 
            "source": "## How would an approach other than classification perform?"
        }, 
        {
            "cell_type": "markdown", 
            "metadata": {}, 
            "source": "### K-means clustering"
        }, 
        {
            "cell_type": "code", 
            "metadata": {
                "scrolled": false
            }, 
            "source": "######## Kmeans clustering, join to the PCA set #########\nfrom numpy import array\nfrom math import sqrt\nfrom pyspark.ml.clustering import KMeans\n\nkmeans = KMeans(k=2,featuresCol='features')\nmodel = kmeans.fit(pca_result)\ncenters = model.clusterCenters()\nkmeans_result = model.transform(pca_result).select('id','features','prediction')\n\nviz_df = pca_result.join(kmeans_result,on=['id','features'],how='inner')\\\n                   .select('id','features','x_coord','y_coord','prediction','label')\n\nviz_df = viz_df.dropna(how='any',subset=['x_coord','y_coord']).toPandas()\n\nsns.lmplot(x='x_coord',y='y_coord',hue='prediction',col='label',data=viz_df,\n           fit_reg=False,palette={1:'red',0:'green'},size=6,aspect=1)", 
            "outputs": [
                {
                    "metadata": {}, 
                    "data": {
                        "text/plain": "<seaborn.axisgrid.FacetGrid at 0x7f1e1830f110>"
                    }, 
                    "execution_count": 49, 
                    "output_type": "execute_result"
                }, 
                {
                    "metadata": {}, 
                    "data": {
                        "text/plain": "<matplotlib.figure.Figure at 0x7f1e1830f750>", 
                        "image/png": "iVBORw0KGgoAAAANSUhEUgAAA50AAAGoCAYAAADFBIF9AAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nAAAAWIakEwAAAABgGZJOAAAAAIBl/j9I0sMGCgrVjgAAAABJRU5ErkJggg==\n"
                    }, 
                    "output_type": "display_data"
                }
            ], 
            "execution_count": 49
        }, 
        {
            "cell_type": "markdown", 
            "metadata": {
                "collapsed": true
            }, 
            "source": "# 5. Deploy to Cloud by using Watson Machine Learning Repo\n\n- Deploy your model with a few lines of code\n- Provides a link that can be embedded into applications (easy hand-off from data scientists to developers)\n\nIBM WML Reference: http://watson-ml-api.mybluemix.net/#/Online32Scoring\n"
        }, 
        {
            "cell_type": "markdown", 
            "metadata": {}, 
            "source": "### Step 1: Setup credentials ###"
        }, 
        {
            "cell_type": "code", 
            "metadata": {
                "collapsed": true
            }, 
            "source": "# The code was removed by DSX for sharing.", 
            "outputs": [], 
            "execution_count": 79
        }, 
        {
            "cell_type": "markdown", 
            "metadata": {}, 
            "source": "### Step 2: Save model ### "
        }, 
        {
            "cell_type": "code", 
            "metadata": {
                "collapsed": true
            }, 
            "source": "### Required imports ###\nimport urllib3, requests, json\n\n### Authenticate ML repo ###\nfrom repository.mlrepositoryclient import MLRepositoryClient\nfrom repository.mlrepositoryartifact import MLRepositoryArtifact\n\nml_repository_client = MLRepositoryClient(service_path)\nml_repository_client.authorize(username, password)\n\n### Save model ###\nmodel_artifact_lr = MLRepositoryArtifact(lr_model, training_data=train, name=\"Logistic Regression Classifer\")\nsaved_model = ml_repository_client.models.save(model_artifact_lr)", 
            "outputs": [], 
            "execution_count": 80
        }, 
        {
            "cell_type": "markdown", 
            "metadata": {}, 
            "source": "### Load saved model and make predictions (batch) ###"
        }, 
        {
            "cell_type": "code", 
            "metadata": {}, 
            "source": "#### load model ###\nloadedModelArtifact = ml_repository_client.models.get(saved_model.uid)\n\n### make predictions ####\npredictions = loadedModelArtifact.model_instance().transform(test)\nprecision_recall(predictions)", 
            "outputs": [
                {
                    "metadata": {}, 
                    "data": {
                        "text/plain": "{'pre_0': 0.9907407407407407,\n 'pre_1': 1.0,\n 'recall_0': 1.0,\n 'recall_1': 0.9836065573770492}"
                    }, 
                    "execution_count": 81, 
                    "output_type": "execute_result"
                }
            ], 
            "execution_count": 81
        }, 
        {
            "cell_type": "markdown", 
            "metadata": {}, 
            "source": "### Obtain token to interact with REST API ###"
        }, 
        {
            "cell_type": "code", 
            "metadata": {
                "collapsed": true
            }, 
            "source": "### get token ###\nheaders = urllib3.util.make_headers(basic_auth='{}:{}'.format(username, password))\nurl = '{}/v2/identity/token'.format(service_path)\nresponse = requests.get(url, headers=headers)\nmltoken = json.loads(response.text).get('token')", 
            "outputs": [], 
            "execution_count": 55
        }, 
        {
            "cell_type": "markdown", 
            "metadata": {}, 
            "source": "### Deploy model ###"
        }, 
        {
            "cell_type": "code", 
            "metadata": {
                "collapsed": true
            }, 
            "source": "### deploy model ###\nendpoint_online = service_path + \"/v2/online/deployments/\"\nheader_online = {'Content-Type': 'application/json', 'Authorization': mltoken}\npayload_online = {\"artifactVersionHref\": saved_model.meta.prop(\"modelVersionHref\"), \"name\": \"Product Line Prediction\"}\n\nresponse_online = requests.post(endpoint_online, json=payload_online, headers=header_online)\nscoring_href = json.loads(response_online.text).get('entity').get('scoringHref') ## this is the endpoint used for online scoring", 
            "outputs": [], 
            "execution_count": 53
        }, 
        {
            "cell_type": "markdown", 
            "metadata": {}, 
            "source": "### Get list of existing models, obtain scoring endpoint for the desired model ###"
        }, 
        {
            "cell_type": "code", 
            "metadata": {
                "collapsed": true
            }, 
            "source": "### get a list of existing deployments ###\nurl = service_path + \"/v2/online/deployments\"\nheader_online = {'Content-Type': 'application/json', 'Authorization': mltoken}\n\n# get scoring endpoint for this model\nresp = requests.get(url,headers=header_online)\nscoring_href = resp.json()['resources'][0]['entity']['scoringHref']", 
            "outputs": [], 
            "execution_count": 114
        }, 
        {
            "cell_type": "markdown", 
            "metadata": {}, 
            "source": "### Get predictions from model (online predictions) ###"
        }, 
        {
            "cell_type": "code", 
            "metadata": {
                "collapsed": true
            }, 
            "source": "### example of getting prediction with online scoring ####\ntest_row = test.take(5)[3]\ntest_features = [test_row[col] for col in test.columns if col not in ['label']]\n\n# get prediction from online scoring endpoint (scoring_href from above)\npayload_scoring = {\"record\":test_features}\nresponse_scoring = requests.put(scoring_href, json=payload_scoring, headers=header_online)\n", 
            "outputs": [], 
            "execution_count": 115
        }, 
        {
            "cell_type": "code", 
            "metadata": {}, 
            "source": "from IPython.display import display\n\noutput = response_scoring.json()\nprint ('Output from scoring endpoint')\ndisplay(output)", 
            "outputs": [
                {
                    "text": "Output from scoring endpoint\n", 
                    "output_type": "stream", 
                    "name": "stdout"
                }, 
                {
                    "metadata": {}, 
                    "data": {
                        "text/plain": "{u'result': {u'AREA_MEAN': 1264.0,\n  u'AREA_SE': 87.87000274658203,\n  u'AREA_WORST': 1750.0,\n  u'COMPACTNESS_MEAN': 0.13130000233650208,\n  u'COMPACTNESS_SE': 0.03482000157237053,\n  u'COMPACTNESS_WORST': 0.3538999855518341,\n  u'CONCAVE POINTS_MEAN': 0.08682999759912491,\n  u'CONCAVE POINTS_SE': 0.012690000236034393,\n  u'CONCAVE POINTS_WORST': 0.15729999542236328,\n  u'CONCAVITY_MEAN': 0.14650000631809235,\n  u'CONCAVITY_SE': 0.042319998145103455,\n  u'CONCAVITY_WORST': 0.4097999930381775,\n  u'FRACTAL_DIMENSION_SE': 0.0044109998270869255,\n  u'FRACTAL_DIMENSION_WORST': 0.08367999643087387,\n  u'ID': 86208,\n  u'PERIMETER_MEAN': 132.39999389648438,\n  u'PERIMETER_SE': 4.553999900817871,\n  u'PERIMETER_WORST': 156.10000610351562,\n  u'RADIUS_MEAN': 20.260000228881836,\n  u'RADIUS_SE': 0.7576000094413757,\n  u'RADIUS_WORST': 24.219999313354492,\n  u'SMOOTHNESS_MEAN': 0.09077999740839005,\n  u'SMOOTHNESS_WORST': 0.11900000274181366,\n  u'SYMMETRY_MEAN': 0.2094999998807907,\n  u'SYMMETRY_WORST': 0.36890000104904175,\n  u'TEXTURE_MEAN': 23.030000686645508,\n  u'TEXTURE_WORST': 31.59000015258789,\n  u'features': {u'values': [20.260000228881836,\n    23.030000686645508,\n    132.39999389648438,\n    1264.0,\n    0.09077999740839005,\n    0.13130000233650208,\n    0.14650000631809235,\n    0.08682999759912491,\n    0.2094999998807907,\n    0.7576000094413757,\n    4.553999900817871,\n    87.87000274658203,\n    0.03482000157237053,\n    0.042319998145103455,\n    0.012690000236034393,\n    0.0044109998270869255,\n    24.219999313354492,\n    31.59000015258789,\n    156.10000610351562,\n    1750.0,\n    0.11900000274181366,\n    0.3538999855518341,\n    0.4097999930381775,\n    0.15729999542236328,\n    0.36890000104904175,\n    0.08367999643087387]},\n  u'prediction': 1.0,\n  u'probability': {u'values': [1.413173287357375e-34, 1.0]},\n  u'rawPrediction': {u'values': [-77.94205542771127, 77.94205542771127]}}}"
                    }, 
                    "output_type": "display_data"
                }
            ], 
            "execution_count": 118
        }, 
        {
            "cell_type": "code", 
            "metadata": {}, 
            "source": "print 'Prediction:',output['result']['prediction']\nprint 'Probability:',output['result']['probability']['values']\nprint 'RawPrediction:',output['result']['probability']['values']", 
            "outputs": [
                {
                    "text": "Prediction: 1.0\nProbability: [1.413173287357375e-34, 1.0]\nRawPrediction: [1.413173287357375e-34, 1.0]\n", 
                    "output_type": "stream", 
                    "name": "stdout"
                }
            ], 
            "execution_count": 121
        }, 
        {
            "cell_type": "code", 
            "metadata": {
                "collapsed": true
            }, 
            "source": "", 
            "outputs": [], 
            "execution_count": null
        }
    ], 
    "metadata": {
        "kernelspec": {
            "display_name": "Python 2 with Spark 2.0", 
            "name": "python2-spark20", 
            "language": "python"
        }, 
        "language_info": {
            "codemirror_mode": {
                "version": 2, 
                "name": "ipython"
            }, 
            "version": "2.7.11", 
            "nbconvert_exporter": "python", 
            "mimetype": "text/x-python", 
            "file_extension": ".py", 
            "name": "python", 
            "pygments_lexer": "ipython2"
        }
    }, 
    "nbformat_minor": 1
}