oversampling-vs-undersampling-techniques.ipynb

{"cells":[{"metadata":{},"cell_type":"markdown","source":"## CLASSIFY HIGHLY UNBALANCED DATA\n\nThis Notebook is a step by step approach to make a classifier on a highly unbalanced dataset with 6 classes in which 77 % of the data falls in Class 1.\n\nI have resampling technique to overcome this unbalanced issue. As you know, there are two ways to handle it either by OVERSAMPLING or UNDERSAMPLING. I have tried both the techniques to see which will perform better.\n\nFor OVERSAMPLING - I used SMOTE works by selecting examples that are close in the feature space, drawing a line between the examples in the feature space and drawing a new sample at a point along that line. \n\nFor UNDERSAMPLING - Undersampling techniques falls under two category - \n\n**\"Select To Keep or Select To Delete\"**\n\nSELECT TO KEEP - Techniques under this category selects some of the values from the majority class applying some algorithms. e.g. NearMiss 1,2,3 and Condensed Nearest Neighbor\n\nSELECT TO DELETE - It applies algorithm (KNN) to delete some of the values from the majority class. e.g. Tomek Links, Edited Nearest Neighbors\n\n****\n\n## CONCLUSION\n\nBecause the data is highly dimensional that is it has lots of feature , in that case, oversampling techniques like SMOTE performs very bad, whereas undersampling techniques performed better. \n\nHence we can conclude SMOTE does not perform well on the multidimensional data."},{"metadata":{"_uuid":"8f2839f25d086af736a60e9eeb907d3b93b6e0e5","_cell_guid":"b1076dfc-b9ad-4769-8c92-a6c4dae69d19","trusted":true},"cell_type":"code","source":"# IMPORT LIBRARIES\nimport pandas as pd\nimport numpy as np\nimport joblib\nimport seaborn as sns\nimport matplotlib.pyplot as plt\nfrom collections import Counter\nfrom imblearn.over_sampling import SMOTE\nfrom imblearn.under_sampling import NearMiss, CondensedNearestNeighbour, EditedNearestNeighbours, TomekLinks, OneSidedSelection, NeighbourhoodCleaningRule\nfrom sklearn.preprocessing import MinMaxScaler\nfrom sklearn.svm import SVC\nfrom sklearn.neighbors import KNeighborsClassifier\nfrom sklearn.tree import DecisionTreeClassifier\nfrom sklearn.pipeline import Pipeline\nfrom sklearn.metrics import fbeta_score, make_scorer\nfrom sklearn.model_selection import train_test_split, KFold, cross_val_score, GridSearchCV,RepeatedStratifiedKFold\nfrom sklearn.ensemble import RandomForestClassifier, AdaBoostClassifier, BaggingClassifier, GradientBoostingClassifier, ExtraTreesClassifier","execution_count":1,"outputs":[]},{"metadata":{"_uuid":"d629ff2d2480ee46fbb7e2d37f6b5fab8052498a","_cell_guid":"79c7e3d0-c299-4dcb-8224-4455121ee9b0","trusted":true},"cell_type":"code","source":"# READ DATA\ntrain_data = pd.read_csv('../input/highly-unbalanced-multiclass6-dataset/spenddata.csv')\ntrain_data = train_data[train_data.columns[1:]]\ntrain_data.head()","execution_count":2,"outputs":[{"output_type":"execute_result","execution_count":2,"data":{"text/plain":"   month  var8  var6  a.1  a.2  a.3  a.4  var5  b.5  b.6  ...  c.276  c.277  \\\n0      1   2.0   NaN    1    5    1   57    34    1  NaN  ...      1      0   \n1      1   2.0   NaN    1    4    1   57    34    2  3.0  ...      0      0   \n2      1   2.0   NaN    1    5    1   57    42    2  1.0  ...      0      0   \n3      1   2.0   NaN    1    6    1   57    34    2  4.0  ...      0      0   \n4      1   2.0   NaN    1    8    1   22     1    1  NaN  ...      0      0   \n\n   c.278  c.279  c.280  c.281  c.282  c.283  f.284  t.158  \n0      0      0      1      0      0      0    5.0    NaN  \n1      0      0      0      0      1      0    NaN    NaN  \n2      0      0      0      0      0      0    3.0    NaN  \n3      0      0      1      0      1      0    5.0    NaN  \n4      0      0      0      0      0      0    5.0    NaN  \n\n[5 rows x 300 columns]","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>month</th>\n      <th>var8</th>\n      <th>var6</th>\n      <th>a.1</th>\n      <th>a.2</th>\n      <th>a.3</th>\n      <th>a.4</th>\n      <th>var5</th>\n      <th>b.5</th>\n      <th>b.6</th>\n      <th>...</th>\n      <th>c.276</th>\n      <th>c.277</th>\n      <th>c.278</th>\n      <th>c.279</th>\n      <th>c.280</th>\n      <th>c.281</th>\n      <th>c.282</th>\n      <th>c.283</th>\n      <th>f.284</th>\n      <th>t.158</th>\n    </tr>\n  </thead>\n  <tbody>\n    <tr>\n      <th>0</th>\n      <td>1</td>\n      <td>2.0</td>\n      <td>NaN</td>\n      <td>1</td>\n      <td>5</td>\n      <td>1</td>\n      <td>57</td>\n      <td>34</td>\n      <td>1</td>\n      <td>NaN</td>\n      <td>...</td>\n      <td>1</td>\n      <td>0</td>\n      <td>0</td>\n      <td>0</td>\n      <td>1</td>\n      <td>0</td>\n      <td>0</td>\n      <td>0</td>\n      <td>5.0</td>\n      <td>NaN</td>\n    </tr>\n    <tr>\n      <th>1</th>\n      <td>1</td>\n      <td>2.0</td>\n      <td>NaN</td>\n      <td>1</td>\n      <td>4</td>\n      <td>1</td>\n      <td>57</td>\n      <td>34</td>\n      <td>2</td>\n      <td>3.0</td>\n      <td>...</td>\n      <td>0</td>\n      <td>0</td>\n      <td>0</td>\n      <td>0</td>\n      <td>0</td>\n      <td>0</td>\n      <td>1</td>\n      <td>0</td>\n      <td>NaN</td>\n      <td>NaN</td>\n    </tr>\n    <tr>\n      <th>2</th>\n      <td>1</td>\n      <td>2.0</td>\n      <td>NaN</td>\n      <td>1</td>\n      <td>5</td>\n      <td>1</td>\n      <td>57</td>\n      <td>42</td>\n      <td>2</td>\n      <td>1.0</td>\n      <td>...</td>\n      <td>0</td>\n      <td>0</td>\n      <td>0</td>\n      <td>0</td>\n      <td>0</td>\n      <td>0</td>\n      <td>0</td>\n      <td>0</td>\n      <td>3.0</td>\n      <td>NaN</td>\n    </tr>\n    <tr>\n      <th>3</th>\n      <td>1</td>\n      <td>2.0</td>\n      <td>NaN</td>\n      <td>1</td>\n      <td>6</td>\n      <td>1</td>\n      <td>57</td>\n      <td>34</td>\n      <td>2</td>\n      <td>4.0</td>\n      <td>...</td>\n      <td>0</td>\n      <td>0</td>\n      <td>0</td>\n      <td>0</td>\n      <td>1</td>\n      <td>0</td>\n      <td>1</td>\n      <td>0</td>\n      <td>5.0</td>\n      <td>NaN</td>\n    </tr>\n    <tr>\n      <th>4</th>\n      <td>1</td>\n      <td>2.0</td>\n      <td>NaN</td>\n      <td>1</td>\n      <td>8</td>\n      <td>1</td>\n      <td>22</td>\n      <td>1</td>\n      <td>1</td>\n      <td>NaN</td>\n      <td>...</td>\n      <td>0</td>\n      <td>0</td>\n      <td>0</td>\n      <td>0</td>\n      <td>0</td>\n      <td>0</td>\n      <td>0</td>\n      <td>0</td>\n      <td>5.0</td>\n      <td>NaN</td>\n    </tr>\n  </tbody>\n</table>\n<p>5 rows × 300 columns</p>\n</div>"},"metadata":{}}]},{"metadata":{"trusted":true},"cell_type":"code","source":"plt.figure(figsize=(15,6))\nsns.countplot(train_data.pov6, palette='Set2')\nplt.show()","execution_count":3,"outputs":[{"output_type":"display_data","data":{"text/plain":"<Figure size 1080x432 with 1 Axes>","image/png":"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\n"},"metadata":{"needs_background":"light"}}]},{"metadata":{"trusted":true},"cell_type":"code","source":"train_data.info()","execution_count":4,"outputs":[{"output_type":"stream","text":"<class 'pandas.core.frame.DataFrame'>\nRangeIndex: 18379 entries, 0 to 18378\nColumns: 300 entries, month to t.158\ndtypes: float64(134), int64(164), object(2)\nmemory usage: 42.1+ MB\n","name":"stdout"}]},{"metadata":{},"cell_type":"markdown","source":"DATA has 18379 DATA POINTS and 300 FEATURES and the TARGET FEATURE is **'pov6'**"},{"metadata":{"trusted":true},"cell_type":"code","source":"train_data.pov6","execution_count":5,"outputs":[{"output_type":"execute_result","execution_count":5,"data":{"text/plain":"0        1\n1        1\n2        1\n3        1\n4        2\n        ..\n18374    1\n18375    1\n18376    5\n18377    1\n18378    2\nName: pov6, Length: 18379, dtype: int64"},"metadata":{}}]},{"metadata":{},"cell_type":"markdown","source":"## TEST DATA"},{"metadata":{"trusted":true},"cell_type":"code","source":"# READ TEST DATA\ntest_data = pd.read_csv('../input/highly-unbalanced-multiclass6-dataset/testdata.csv')\ntest_data = test_data[test_data.columns[1:]]\ntest_data.info()","execution_count":6,"outputs":[{"output_type":"stream","text":"<class 'pandas.core.frame.DataFrame'>\nRangeIndex: 4595 entries, 0 to 4594\nColumns: 299 entries, month to f.284\ndtypes: float64(127), int64(170), object(2)\nmemory usage: 10.5+ MB\n","name":"stdout"}]},{"metadata":{},"cell_type":"markdown","source":"## DROP DUPLICATES"},{"metadata":{"trusted":true},"cell_type":"code","source":"train_data.drop_duplicates()","execution_count":7,"outputs":[{"output_type":"execute_result","execution_count":7,"data":{"text/plain":"       month  var8  var6  a.1  a.2  a.3  a.4  var5  b.5  b.6  ...  c.276  \\\n0          1   2.0   NaN    1    5    1   57    34    1  NaN  ...      1   \n1          1   2.0   NaN    1    4    1   57    34    2  3.0  ...      0   \n2          1   2.0   NaN    1    5    1   57    42    2  1.0  ...      0   \n3          1   2.0   NaN    1    6    1   57    34    2  4.0  ...      0   \n4          1   2.0   NaN    1    8    1   22     1    1  NaN  ...      0   \n...      ...   ...   ...  ...  ...  ...  ...   ...  ...  ...  ...    ...   \n18374     12   NaN   NaN    1    3    1   58    20    2  2.0  ...      0   \n18375     12   NaN   NaN    1    3    1  123    31    1  NaN  ...      0   \n18376     12   NaN   NaN    1    5    1   44    21    2  1.0  ...      0   \n18377     12   NaN   NaN    1    4    1   44    21    1  NaN  ...      0   \n18378     12   3.0   NaN    1    6    1   10    33    1  NaN  ...      0   \n\n       c.277  c.278  c.279  c.280  c.281  c.282  c.283  f.284  t.158  \n0          0      0      0      1      0      0      0    5.0    NaN  \n1          0      0      0      0      0      1      0    NaN    NaN  \n2          0      0      0      0      0      0      0    3.0    NaN  \n3          0      0      0      1      0      1      0    5.0    NaN  \n4          0      0      0      0      0      0      0    5.0    NaN  \n...      ...    ...    ...    ...    ...    ...    ...    ...    ...  \n18374      0      0      0      0      0      0      0    1.0    NaN  \n18375      0      0      0      0      0      0      0    2.0    NaN  \n18376      0      0      0      0      0      0      0    1.0    NaN  \n18377      0      1      0      0      0      1      0    1.0    NaN  \n18378      0      0      0      1      0      0      0    NaN    NaN  \n\n[18379 rows x 300 columns]","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>month</th>\n      <th>var8</th>\n      <th>var6</th>\n      <th>a.1</th>\n      <th>a.2</th>\n      <th>a.3</th>\n      <th>a.4</th>\n      <th>var5</th>\n      <th>b.5</th>\n      <th>b.6</th>\n      <th>...</th>\n      <th>c.276</th>\n      <th>c.277</th>\n      <th>c.278</th>\n      <th>c.279</th>\n      <th>c.280</th>\n      <th>c.281</th>\n      <th>c.282</th>\n      <th>c.283</th>\n      <th>f.284</th>\n      <th>t.158</th>\n    </tr>\n  </thead>\n  <tbody>\n    <tr>\n      <th>0</th>\n      <td>1</td>\n      <td>2.0</td>\n      <td>NaN</td>\n      <td>1</td>\n      <td>5</td>\n      <td>1</td>\n      <td>57</td>\n      <td>34</td>\n      <td>1</td>\n      <td>NaN</td>\n      <td>...</td>\n      <td>1</td>\n      <td>0</td>\n      <td>0</td>\n      <td>0</td>\n      <td>1</td>\n      <td>0</td>\n      <td>0</td>\n      <td>0</td>\n      <td>5.0</td>\n      <td>NaN</td>\n    </tr>\n    <tr>\n      <th>1</th>\n      <td>1</td>\n      <td>2.0</td>\n      <td>NaN</td>\n      <td>1</td>\n      <td>4</td>\n      <td>1</td>\n      <td>57</td>\n      <td>34</td>\n      <td>2</td>\n      <td>3.0</td>\n      <td>...</td>\n      <td>0</td>\n      <td>0</td>\n      <td>0</td>\n      <td>0</td>\n      <td>0</td>\n      <td>0</td>\n      <td>1</td>\n      <td>0</td>\n      <td>NaN</td>\n      <td>NaN</td>\n    </tr>\n    <tr>\n      <th>2</th>\n      <td>1</td>\n      <td>2.0</td>\n      <td>NaN</td>\n      <td>1</td>\n      <td>5</td>\n      <td>1</td>\n      <td>57</td>\n      <td>42</td>\n      <td>2</td>\n      <td>1.0</td>\n      <td>...</td>\n      <td>0</td>\n      <td>0</td>\n      <td>0</td>\n      <td>0</td>\n      <td>0</td>\n      <td>0</td>\n      <td>0</td>\n      <td>0</td>\n      <td>3.0</td>\n      <td>NaN</td>\n    </tr>\n    <tr>\n      <th>3</th>\n      <td>1</td>\n      <td>2.0</td>\n      <td>NaN</td>\n      <td>1</td>\n      <td>6</td>\n      <td>1</td>\n      <td>57</td>\n      <td>34</td>\n      <td>2</td>\n      <td>4.0</td>\n      <td>...</td>\n      <td>0</td>\n      <td>0</td>\n      <td>0</td>\n      <td>0</td>\n      <td>1</td>\n      <td>0</td>\n      <td>1</td>\n      <td>0</td>\n      <td>5.0</td>\n      <td>NaN</td>\n    </tr>\n    <tr>\n      <th>4</th>\n      <td>1</td>\n      <td>2.0</td>\n      <td>NaN</td>\n      <td>1</td>\n      <td>8</td>\n      <td>1</td>\n      <td>22</td>\n      <td>1</td>\n      <td>1</td>\n      <td>NaN</td>\n      <td>...</td>\n      <td>0</td>\n      <td>0</td>\n      <td>0</td>\n      <td>0</td>\n      <td>0</td>\n      <td>0</td>\n      <td>0</td>\n      <td>0</td>\n      <td>5.0</td>\n      <td>NaN</td>\n    </tr>\n    <tr>\n      <th>...</th>\n      <td>...</td>\n      <td>...</td>\n      <td>...</td>\n      <td>...</td>\n      <td>...</td>\n      <td>...</td>\n      <td>...</td>\n      <td>...</td>\n      <td>...</td>\n      <td>...</td>\n      <td>...</td>\n      <td>...</td>\n      <td>...</td>\n      <td>...</td>\n      <td>...</td>\n      <td>...</td>\n      <td>...</td>\n      <td>...</td>\n      <td>...</td>\n      <td>...</td>\n      <td>...</td>\n    </tr>\n    <tr>\n      <th>18374</th>\n      <td>12</td>\n      <td>NaN</td>\n      <td>NaN</td>\n      <td>1</td>\n      <td>3</td>\n      <td>1</td>\n      <td>58</td>\n      <td>20</td>\n      <td>2</td>\n      <td>2.0</td>\n      <td>...</td>\n      <td>0</td>\n      <td>0</td>\n      <td>0</td>\n      <td>0</td>\n      <td>0</td>\n      <td>0</td>\n      <td>0</td>\n      <td>0</td>\n      <td>1.0</td>\n      <td>NaN</td>\n    </tr>\n    <tr>\n      <th>18375</th>\n      <td>12</td>\n      <td>NaN</td>\n      <td>NaN</td>\n      <td>1</td>\n      <td>3</td>\n      <td>1</td>\n      <td>123</td>\n      <td>31</td>\n      <td>1</td>\n      <td>NaN</td>\n      <td>...</td>\n      <td>0</td>\n      <td>0</td>\n      <td>0</td>\n      <td>0</td>\n      <td>0</td>\n      <td>0</td>\n      <td>0</td>\n      <td>0</td>\n      <td>2.0</td>\n      <td>NaN</td>\n    </tr>\n    <tr>\n      <th>18376</th>\n      <td>12</td>\n      <td>NaN</td>\n      <td>NaN</td>\n      <td>1</td>\n      <td>5</td>\n      <td>1</td>\n      <td>44</td>\n      <td>21</td>\n      <td>2</td>\n      <td>1.0</td>\n      <td>...</td>\n      <td>0</td>\n      <td>0</td>\n      <td>0</td>\n      <td>0</td>\n      <td>0</td>\n      <td>0</td>\n      <td>0</td>\n      <td>0</td>\n      <td>1.0</td>\n      <td>NaN</td>\n    </tr>\n    <tr>\n      <th>18377</th>\n      <td>12</td>\n      <td>NaN</td>\n      <td>NaN</td>\n      <td>1</td>\n      <td>4</td>\n      <td>1</td>\n      <td>44</td>\n      <td>21</td>\n      <td>1</td>\n      <td>NaN</td>\n      <td>...</td>\n      <td>0</td>\n      <td>0</td>\n      <td>1</td>\n      <td>0</td>\n      <td>0</td>\n      <td>0</td>\n      <td>1</td>\n      <td>0</td>\n      <td>1.0</td>\n      <td>NaN</td>\n    </tr>\n    <tr>\n      <th>18378</th>\n      <td>12</td>\n      <td>3.0</td>\n      <td>NaN</td>\n      <td>1</td>\n      <td>6</td>\n      <td>1</td>\n      <td>10</td>\n      <td>33</td>\n      <td>1</td>\n      <td>NaN</td>\n      <td>...</td>\n      <td>0</td>\n      <td>0</td>\n      <td>0</td>\n      <td>0</td>\n      <td>1</td>\n      <td>0</td>\n      <td>0</td>\n      <td>0</td>\n      <td>NaN</td>\n      <td>NaN</td>\n    </tr>\n  </tbody>\n</table>\n<p>18379 rows × 300 columns</p>\n</div>"},"metadata":{}}]},{"metadata":{},"cell_type":"markdown","source":"#### NO DUPLICATES FOUND "},{"metadata":{},"cell_type":"markdown","source":"# REMOVE NULL VALUES\n\n\nWe will make a dictionary with column as KEY and total number of null values in that column as VALUE. Then we will sort that dictionary in descending order by VALUE and remove those columns that have more than 11000 null values."},{"metadata":{"trusted":true},"cell_type":"code","source":"# STORE ALL THE COLUMNS THAT HAS NULL VALUES \ntrain_null_columns = []\n\n# ITERATE THROUGH TRAINING DATA COLUMNS AND CHECK WHETHER IT HAS NULL VALUES OR NOT\nfor index, rows in pd.DataFrame(train_data.isna().any()).iterrows():\n    if rows[0] == True:\n        # COLUMNS WITH NULL VALUES MORE THAN 11000\n        if train_data[index].isna().sum() >= 11000:\n            train_null_columns.append(index)\n        \ntrain_null_columns","execution_count":8,"outputs":[{"output_type":"execute_result","execution_count":8,"data":{"text/plain":"['var6',\n 'b.10',\n 'b.11',\n 'b.13',\n 'b.14',\n 'b.20',\n 'b.21',\n 'c.31',\n 'c.36',\n 'c.37',\n 'c.61',\n 'c.67',\n 'f.105',\n 'f.106',\n 'f.107',\n 'f.108',\n 'f.109',\n 'f.110',\n 'f.111',\n 'f.112',\n 'f.113',\n 'f.114',\n 'f.115',\n 'f.116',\n 'f.117',\n 'f.118',\n 'f.119',\n 'f.120',\n 'f.121',\n 'f.122',\n 'f.123',\n 'a.184',\n 'a.185',\n 'b.195',\n 'c.215',\n 'c.216',\n 'c.217',\n 'c.218',\n 'c.219',\n 'c.220',\n 'c.221',\n 'c.222',\n 'c.223',\n 't.158']"},"metadata":{}}]},{"metadata":{},"cell_type":"markdown","source":"## SIMILARLY CHECK FOR THE NULL VALUES IN TEST DATA"},{"metadata":{},"cell_type":"markdown","source":"As test data has 4595 datapoints we will check for the columns that has null values more than 4000"},{"metadata":{"trusted":true},"cell_type":"code","source":"# STORE ALL THE COLUMNS THAT HAS NULL VALUES \ntest_null_columns = []\n\n# ITERATE THROUGH TEST DATA COLUMNS AND CHECK WHETHER IT HAS NULL VALUES OR NOT\nfor index, rows in pd.DataFrame(test_data.isna().any()).iterrows():\n    if rows[0] == True:\n        # COLUMNS WITH NULL VALUES MORE THAN 11000\n        if test_data[index].isna().sum() >= 4000:\n            test_null_columns.append(index)\n        \ntest_null_columns    ","execution_count":9,"outputs":[{"output_type":"execute_result","execution_count":9,"data":{"text/plain":"['var6',\n 'b.10',\n 'b.11',\n 'pov6',\n 'c.36',\n 'c.37',\n 'c.61',\n 'c.67',\n 'f.105',\n 'f.106',\n 'f.107',\n 'f.108',\n 'f.109',\n 'f.110',\n 'f.111',\n 'f.114',\n 'f.115',\n 'f.116',\n 'f.117',\n 'f.118',\n 'f.119',\n 'f.120',\n 'f.121',\n 'f.122',\n 'f.123',\n 'a.184',\n 'a.185',\n 'totshopping.rep',\n 'c.215',\n 'c.216',\n 'c.217',\n 'c.218',\n 'c.219',\n 'c.220',\n 'c.221',\n 'c.223']"},"metadata":{}}]},{"metadata":{},"cell_type":"markdown","source":"## COMMON NULL COLUMNS TO BE DELETED"},{"metadata":{"trusted":true},"cell_type":"code","source":"# COLUMNS THAT HAVE MOSTLY NULL VALUES IN TRAIN DATA BUT NOT IN TEST DATA\nset(train_null_columns) - set(test_null_columns)","execution_count":10,"outputs":[{"output_type":"execute_result","execution_count":10,"data":{"text/plain":"{'b.13',\n 'b.14',\n 'b.195',\n 'b.20',\n 'b.21',\n 'c.222',\n 'c.31',\n 'f.112',\n 'f.113',\n 't.158'}"},"metadata":{}}]},{"metadata":{"trusted":true},"cell_type":"code","source":"# COLUMNS THAT HAVE MOSTLY NULL VALUES TEST DATA BUT NOT IN TRAIN DATA\nset(test_null_columns) - set(train_null_columns)","execution_count":11,"outputs":[{"output_type":"execute_result","execution_count":11,"data":{"text/plain":"{'pov6', 'totshopping.rep'}"},"metadata":{}}]},{"metadata":{},"cell_type":"markdown","source":"As we can see columns **'pov6' and 'totshopping.rep'** are having most of the null values in test data. 'pov6' is our label or target columns so we can delete this column from the test data and after storing the values of column 'pov6' in another variable named \"label\", we can delete it from the training data as well.\n\nLets analyse **'totshopping.rep'** whether we should delete it or not. "},{"metadata":{"trusted":true},"cell_type":"code","source":"test_data['totshopping.rep'].isna().sum()","execution_count":12,"outputs":[{"output_type":"execute_result","execution_count":12,"data":{"text/plain":"4595"},"metadata":{}}]},{"metadata":{"trusted":true},"cell_type":"code","source":"train_data['totshopping.rep'].isna().sum()","execution_count":13,"outputs":[{"output_type":"execute_result","execution_count":13,"data":{"text/plain":"0"},"metadata":{}}]},{"metadata":{},"cell_type":"markdown","source":"So here we can see that column **'totshopping.rep'** has all the values null in test data but no values null in the training data. Then the question is \"WE SHOULD DELETE IT OR NOT\"?\n\nANSWER IS \"YES\", because test data is what on we need to test our model and if the features are not present in the test data, it is totally irrelevant to include it in the training data.\n\n****\nSo the TOTAL NULL COLUMNS TO BE DELETED are \n\n    train_null_columns + set(test_null_columns) - set(train_null_columns)\n    \n    i.e. train_null_columns + {'pov6', 'totshopping.rep'}\n****"},{"metadata":{"trusted":true},"cell_type":"code","source":"# TOTAL NULL COLUMNS TO BE DELETED\ntot_cols = train_null_columns + list(set(test_null_columns) - set(train_null_columns))\n\nprint('TOTAL COLUMNS TO BE DELETED', len(tot_cols))\n\n# DELETE THE NULL VALUED COLUMNS FROM TRAINING DATA \nfiltered_data = train_data.drop(tot_cols, axis=1)\n\nprint('COLUMNS BEFORE', len(train_data.columns))\nprint('COLUMNS AFTER REMOVAL', len(filtered_data.columns))","execution_count":14,"outputs":[{"output_type":"stream","text":"TOTAL COLUMNS TO BE DELETED 46\nCOLUMNS BEFORE 300\nCOLUMNS AFTER REMOVAL 254\n","name":"stdout"}]},{"metadata":{},"cell_type":"markdown","source":"It has removed 46 columns with most of the null values. It has reduced features from 300 to 254."},{"metadata":{"trusted":true},"cell_type":"code","source":"# DELETE THE NULL VALUED COLUMNS FROM TEST DATA \nfiltered_test_data = test_data.copy()\n\nfor col in tot_cols:\n    try:\n        filtered_test_data.drop([col], axis=1, inplace=True)\n    except:\n        print('{} COLUMN not in TEST DATA '.format(col))\n\nprint()\nprint('COLUMNS BEFORE', len(test_data.columns))\nprint('COLUMNS AFTER REMOVAL', len(filtered_test_data.columns))","execution_count":15,"outputs":[{"output_type":"stream","text":"t.158 COLUMN not in TEST DATA \n\nCOLUMNS BEFORE 299\nCOLUMNS AFTER REMOVAL 254\n","name":"stdout"}]},{"metadata":{},"cell_type":"markdown","source":"Similarly it has reduced test data features from 299 to 254"},{"metadata":{},"cell_type":"markdown","source":"## CHECK WHETHER TEST DATA AND TRAINING DATA BOTH HAVE SAME COLUMNS OR NOT"},{"metadata":{"trusted":true},"cell_type":"code","source":"if set(filtered_data.columns.tolist()) == set(filtered_test_data.columns.tolist()):\n    print('THEY BOTH HAVE SAME COLUMNS')\nelse:\n    print(set(filtered_data.columns.tolist()) - set(filtered_test_data.columns.tolist()))\n    print(set(filtered_test_data.columns.tolist()) - set(filtered_data.columns.tolist()))","execution_count":16,"outputs":[{"output_type":"stream","text":"THEY BOTH HAVE SAME COLUMNS\n","name":"stdout"}]},{"metadata":{},"cell_type":"markdown","source":"## MAKE LABEL\n\n'pov6' column as LABEL data \n\n'filtered_data' is our FEATURE data."},{"metadata":{"trusted":true},"cell_type":"code","source":"# MAKE LABEL\nlabel = pd.DataFrame(train_data.pov6)","execution_count":17,"outputs":[]},{"metadata":{},"cell_type":"markdown","source":"## STRING FEATURE VALUES\n\nList all the features columns that has string values."},{"metadata":{"trusted":true},"cell_type":"code","source":"# SELECT COLUMNS WITH OBJECT DATATYPE\nfiltered_data.columns[filtered_data.dtypes=='object']","execution_count":18,"outputs":[{"output_type":"execute_result","execution_count":18,"data":{"text/plain":"Index(['var9', 'respondent.id'], dtype='object')"},"metadata":{}}]},{"metadata":{"trusted":true},"cell_type":"code","source":"filtered_data['var9'].head()","execution_count":19,"outputs":[{"output_type":"execute_result","execution_count":19,"data":{"text/plain":"0     Mono\n1     Mono\n2    Multi\n3     Mono\n4    Multi\nName: var9, dtype: object"},"metadata":{}}]},{"metadata":{"trusted":true},"cell_type":"code","source":"filtered_data['respondent.id'].head()","execution_count":20,"outputs":[{"output_type":"execute_result","execution_count":20,"data":{"text/plain":"0    2016_1\n1    2016_2\n2    2016_3\n3    2016_4\n4    2016_5\nName: respondent.id, dtype: object"},"metadata":{}}]},{"metadata":{},"cell_type":"markdown","source":"As we can see column **respondent.id** is not of any use, so we can delete it. We need to encode column **var9** to numerical values, we can do one hot encoding using **pandas.get_dummies**"},{"metadata":{"trusted":true},"cell_type":"code","source":"# ENCODE STRING COLUMNS \npd.get_dummies(filtered_data.var9)","execution_count":21,"outputs":[{"output_type":"execute_result","execution_count":21,"data":{"text/plain":"       Mono  Multi\n0         1      0\n1         1      0\n2         0      1\n3         1      0\n4         0      1\n...     ...    ...\n18374     1      0\n18375     1      0\n18376     1      0\n18377     1      0\n18378     1      0\n\n[18379 rows x 2 columns]","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>Mono</th>\n      <th>Multi</th>\n    </tr>\n  </thead>\n  <tbody>\n    <tr>\n      <th>0</th>\n      <td>1</td>\n      <td>0</td>\n    </tr>\n    <tr>\n      <th>1</th>\n      <td>1</td>\n      <td>0</td>\n    </tr>\n    <tr>\n      <th>2</th>\n      <td>0</td>\n      <td>1</td>\n    </tr>\n    <tr>\n      <th>3</th>\n      <td>1</td>\n      <td>0</td>\n    </tr>\n    <tr>\n      <th>4</th>\n      <td>0</td>\n      <td>1</td>\n    </tr>\n    <tr>\n      <th>...</th>\n      <td>...</td>\n      <td>...</td>\n    </tr>\n    <tr>\n      <th>18374</th>\n      <td>1</td>\n      <td>0</td>\n    </tr>\n    <tr>\n      <th>18375</th>\n      <td>1</td>\n      <td>0</td>\n    </tr>\n    <tr>\n      <th>18376</th>\n      <td>1</td>\n      <td>0</td>\n    </tr>\n    <tr>\n      <th>18377</th>\n      <td>1</td>\n      <td>0</td>\n    </tr>\n    <tr>\n      <th>18378</th>\n      <td>1</td>\n      <td>0</td>\n    </tr>\n  </tbody>\n</table>\n<p>18379 rows × 2 columns</p>\n</div>"},"metadata":{}}]},{"metadata":{},"cell_type":"markdown","source":"## CONCATENATE THE ENCODED DATA WITH TRAINING DATA \n\nAfter concatenation , remove both the string columns"},{"metadata":{"trusted":true},"cell_type":"code","source":"# CONCAT THE DUMMY DATA WITH THE TRAINING DATA\nfiltered_data = pd.concat([filtered_data, pd.get_dummies(filtered_data.var9)], axis=1 )\n\n# REMOVE UNWANTED STRING COLUMN\nfiltered_data.drop(['respondent.id','var9'], axis=1, inplace=True)\nlen(filtered_data.columns)","execution_count":22,"outputs":[{"output_type":"execute_result","execution_count":22,"data":{"text/plain":"254"},"metadata":{}}]},{"metadata":{},"cell_type":"markdown","source":"## SAME FOR TEST DATA"},{"metadata":{"trusted":true},"cell_type":"code","source":"# CONCAT THE DUMMY DATA WITH THE TEST DATA\nfiltered_test_data = pd.concat([filtered_test_data, pd.get_dummies(filtered_test_data.var9)], axis=1 )\n\n# REMOVE UNWANTED STRING COLUMN\nfiltered_test_data.drop(['respondent.id','var9'], axis=1, inplace=True)\nlen(filtered_test_data.columns)","execution_count":23,"outputs":[{"output_type":"execute_result","execution_count":23,"data":{"text/plain":"254"},"metadata":{}}]},{"metadata":{},"cell_type":"markdown","source":"## FILL NULL VALUES WITH MEAN\n\nFill the null values of each column with mean value of that particular column."},{"metadata":{"trusted":true},"cell_type":"code","source":"# BEFORE REMOVING NULL VALUES\nfiltered_data.head()","execution_count":24,"outputs":[{"output_type":"execute_result","execution_count":24,"data":{"text/plain":"   month  var8  a.1  a.2  a.3  a.4  var5  b.5  b.6  b.7  ...  c.277  c.278  \\\n0      1   2.0    1    5    1   57    34    1  NaN    1  ...      0      0   \n1      1   2.0    1    4    1   57    34    2  3.0    1  ...      0      0   \n2      1   2.0    1    5    1   57    42    2  1.0    1  ...      0      0   \n3      1   2.0    1    6    1   57    34    2  4.0    1  ...      0      0   \n4      1   2.0    1    8    1   22     1    1  NaN    1  ...      0      0   \n\n   c.279  c.280  c.281  c.282  c.283  f.284  Mono  Multi  \n0      0      1      0      0      0    5.0     1      0  \n1      0      0      0      1      0    NaN     1      0  \n2      0      0      0      0      0    3.0     0      1  \n3      0      1      0      1      0    5.0     1      0  \n4      0      0      0      0      0    5.0     0      1  \n\n[5 rows x 254 columns]","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>month</th>\n      <th>var8</th>\n      <th>a.1</th>\n      <th>a.2</th>\n      <th>a.3</th>\n      <th>a.4</th>\n      <th>var5</th>\n      <th>b.5</th>\n      <th>b.6</th>\n      <th>b.7</th>\n      <th>...</th>\n      <th>c.277</th>\n      <th>c.278</th>\n      <th>c.279</th>\n      <th>c.280</th>\n      <th>c.281</th>\n      <th>c.282</th>\n      <th>c.283</th>\n      <th>f.284</th>\n      <th>Mono</th>\n      <th>Multi</th>\n    </tr>\n  </thead>\n  <tbody>\n    <tr>\n      <th>0</th>\n      <td>1</td>\n      <td>2.0</td>\n      <td>1</td>\n      <td>5</td>\n      <td>1</td>\n      <td>57</td>\n      <td>34</td>\n      <td>1</td>\n      <td>NaN</td>\n      <td>1</td>\n      <td>...</td>\n      <td>0</td>\n      <td>0</td>\n      <td>0</td>\n      <td>1</td>\n      <td>0</td>\n      <td>0</td>\n      <td>0</td>\n      <td>5.0</td>\n      <td>1</td>\n      <td>0</td>\n    </tr>\n    <tr>\n      <th>1</th>\n      <td>1</td>\n      <td>2.0</td>\n      <td>1</td>\n      <td>4</td>\n      <td>1</td>\n      <td>57</td>\n      <td>34</td>\n      <td>2</td>\n      <td>3.0</td>\n      <td>1</td>\n      <td>...</td>\n      <td>0</td>\n      <td>0</td>\n      <td>0</td>\n      <td>0</td>\n      <td>0</td>\n      <td>1</td>\n      <td>0</td>\n      <td>NaN</td>\n      <td>1</td>\n      <td>0</td>\n    </tr>\n    <tr>\n      <th>2</th>\n      <td>1</td>\n      <td>2.0</td>\n      <td>1</td>\n      <td>5</td>\n      <td>1</td>\n      <td>57</td>\n      <td>42</td>\n      <td>2</td>\n      <td>1.0</td>\n      <td>1</td>\n      <td>...</td>\n      <td>0</td>\n      <td>0</td>\n      <td>0</td>\n      <td>0</td>\n      <td>0</td>\n      <td>0</td>\n      <td>0</td>\n      <td>3.0</td>\n      <td>0</td>\n      <td>1</td>\n    </tr>\n    <tr>\n      <th>3</th>\n      <td>1</td>\n      <td>2.0</td>\n      <td>1</td>\n      <td>6</td>\n      <td>1</td>\n      <td>57</td>\n      <td>34</td>\n      <td>2</td>\n      <td>4.0</td>\n      <td>1</td>\n      <td>...</td>\n      <td>0</td>\n      <td>0</td>\n      <td>0</td>\n      <td>1</td>\n      <td>0</td>\n      <td>1</td>\n      <td>0</td>\n      <td>5.0</td>\n      <td>1</td>\n      <td>0</td>\n    </tr>\n    <tr>\n      <th>4</th>\n      <td>1</td>\n      <td>2.0</td>\n      <td>1</td>\n      <td>8</td>\n      <td>1</td>\n      <td>22</td>\n      <td>1</td>\n      <td>1</td>\n      <td>NaN</td>\n      <td>1</td>\n      <td>...</td>\n      <td>0</td>\n      <td>0</td>\n      <td>0</td>\n      <td>0</td>\n      <td>0</td>\n      <td>0</td>\n      <td>0</td>\n      <td>5.0</td>\n      <td>0</td>\n      <td>1</td>\n    </tr>\n  </tbody>\n</table>\n<p>5 rows × 254 columns</p>\n</div>"},"metadata":{}}]},{"metadata":{"trusted":true},"cell_type":"code","source":"# FILL ALL THE NULL VALUES WITH MEAN\nfiltered_data.fillna(filtered_data.mean(), inplace=True)\nfiltered_data.head()","execution_count":25,"outputs":[{"output_type":"execute_result","execution_count":25,"data":{"text/plain":"   month  var8  a.1  a.2  a.3  a.4  var5  b.5       b.6  b.7  ...  c.277  \\\n0      1   2.0    1    5    1   57    34    1  2.224272    1  ...      0   \n1      1   2.0    1    4    1   57    34    2  3.000000    1  ...      0   \n2      1   2.0    1    5    1   57    42    2  1.000000    1  ...      0   \n3      1   2.0    1    6    1   57    34    2  4.000000    1  ...      0   \n4      1   2.0    1    8    1   22     1    1  2.224272    1  ...      0   \n\n   c.278  c.279  c.280  c.281  c.282  c.283     f.284  Mono  Multi  \n0      0      0      1      0      0      0  5.000000     1      0  \n1      0      0      0      0      1      0  3.038536     1      0  \n2      0      0      0      0      0      0  3.000000     0      1  \n3      0      0      1      0      1      0  5.000000     1      0  \n4      0      0      0      0      0      0  5.000000     0      1  \n\n[5 rows x 254 columns]","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>month</th>\n      <th>var8</th>\n      <th>a.1</th>\n      <th>a.2</th>\n      <th>a.3</th>\n      <th>a.4</th>\n      <th>var5</th>\n      <th>b.5</th>\n      <th>b.6</th>\n      <th>b.7</th>\n      <th>...</th>\n      <th>c.277</th>\n      <th>c.278</th>\n      <th>c.279</th>\n      <th>c.280</th>\n      <th>c.281</th>\n      <th>c.282</th>\n      <th>c.283</th>\n      <th>f.284</th>\n      <th>Mono</th>\n      <th>Multi</th>\n    </tr>\n  </thead>\n  <tbody>\n    <tr>\n      <th>0</th>\n      <td>1</td>\n      <td>2.0</td>\n      <td>1</td>\n      <td>5</td>\n      <td>1</td>\n      <td>57</td>\n      <td>34</td>\n      <td>1</td>\n      <td>2.224272</td>\n      <td>1</td>\n      <td>...</td>\n      <td>0</td>\n      <td>0</td>\n      <td>0</td>\n      <td>1</td>\n      <td>0</td>\n      <td>0</td>\n      <td>0</td>\n      <td>5.000000</td>\n      <td>1</td>\n      <td>0</td>\n    </tr>\n    <tr>\n      <th>1</th>\n      <td>1</td>\n      <td>2.0</td>\n      <td>1</td>\n      <td>4</td>\n      <td>1</td>\n      <td>57</td>\n      <td>34</td>\n      <td>2</td>\n      <td>3.000000</td>\n      <td>1</td>\n      <td>...</td>\n      <td>0</td>\n      <td>0</td>\n      <td>0</td>\n      <td>0</td>\n      <td>0</td>\n      <td>1</td>\n      <td>0</td>\n      <td>3.038536</td>\n      <td>1</td>\n      <td>0</td>\n    </tr>\n    <tr>\n      <th>2</th>\n      <td>1</td>\n      <td>2.0</td>\n      <td>1</td>\n      <td>5</td>\n      <td>1</td>\n      <td>57</td>\n      <td>42</td>\n      <td>2</td>\n      <td>1.000000</td>\n      <td>1</td>\n      <td>...</td>\n      <td>0</td>\n      <td>0</td>\n      <td>0</td>\n      <td>0</td>\n      <td>0</td>\n      <td>0</td>\n      <td>0</td>\n      <td>3.000000</td>\n      <td>0</td>\n      <td>1</td>\n    </tr>\n    <tr>\n      <th>3</th>\n      <td>1</td>\n      <td>2.0</td>\n      <td>1</td>\n      <td>6</td>\n      <td>1</td>\n      <td>57</td>\n      <td>34</td>\n      <td>2</td>\n      <td>4.000000</td>\n      <td>1</td>\n      <td>...</td>\n      <td>0</td>\n      <td>0</td>\n      <td>0</td>\n      <td>1</td>\n      <td>0</td>\n      <td>1</td>\n      <td>0</td>\n      <td>5.000000</td>\n      <td>1</td>\n      <td>0</td>\n    </tr>\n    <tr>\n      <th>4</th>\n      <td>1</td>\n      <td>2.0</td>\n      <td>1</td>\n      <td>8</td>\n      <td>1</td>\n      <td>22</td>\n      <td>1</td>\n      <td>1</td>\n      <td>2.224272</td>\n      <td>1</td>\n      <td>...</td>\n      <td>0</td>\n      <td>0</td>\n      <td>0</td>\n      <td>0</td>\n      <td>0</td>\n      <td>0</td>\n      <td>0</td>\n      <td>5.000000</td>\n      <td>0</td>\n      <td>1</td>\n    </tr>\n  </tbody>\n</table>\n<p>5 rows × 254 columns</p>\n</div>"},"metadata":{}}]},{"metadata":{},"cell_type":"markdown","source":"## SAME FOR THE TEST DATA"},{"metadata":{"trusted":true},"cell_type":"code","source":"# FILL ALL THE NULL VALUES WITH MEAN\nfiltered_test_data.fillna(filtered_test_data.mean(), inplace=True)\nfiltered_test_data.head()","execution_count":26,"outputs":[{"output_type":"execute_result","execution_count":26,"data":{"text/plain":"   month  var8  a.1  a.2  a.3  a.4  var5  b.5       b.6  b.7  ...  c.277  \\\n0      1   1.0    1   11    1   47     8    2  1.000000    1  ...      0   \n1      1   2.0    1   11    1   65    27    1  2.247994    1  ...      0   \n2      1   2.0    1    6    1   65    27    2  1.000000    1  ...      0   \n3      1   1.0    1    3    1    3     2    1  2.247994    1  ...      0   \n4      1   1.0    1    5    1   83     3    1  2.247994    1  ...      0   \n\n   c.278  c.279  c.280  c.281  c.282  c.283     f.284  Mono  Multi  \n0      0      0      0      0      1      0  3.000000     0      0  \n1      0      0      0      0      0      0  3.000000     1      0  \n2      0      0      0      0      1      0  3.000000     1      0  \n3      0      0      0      0      0      0  3.079814     0      0  \n4      0      0      1      0      0      0  3.079814     0      1  \n\n[5 rows x 254 columns]","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>month</th>\n      <th>var8</th>\n      <th>a.1</th>\n      <th>a.2</th>\n      <th>a.3</th>\n      <th>a.4</th>\n      <th>var5</th>\n      <th>b.5</th>\n      <th>b.6</th>\n      <th>b.7</th>\n      <th>...</th>\n      <th>c.277</th>\n      <th>c.278</th>\n      <th>c.279</th>\n      <th>c.280</th>\n      <th>c.281</th>\n      <th>c.282</th>\n      <th>c.283</th>\n      <th>f.284</th>\n      <th>Mono</th>\n      <th>Multi</th>\n    </tr>\n  </thead>\n  <tbody>\n    <tr>\n      <th>0</th>\n      <td>1</td>\n      <td>1.0</td>\n      <td>1</td>\n      <td>11</td>\n      <td>1</td>\n      <td>47</td>\n      <td>8</td>\n      <td>2</td>\n      <td>1.000000</td>\n      <td>1</td>\n      <td>...</td>\n      <td>0</td>\n      <td>0</td>\n      <td>0</td>\n      <td>0</td>\n      <td>0</td>\n      <td>1</td>\n      <td>0</td>\n      <td>3.000000</td>\n      <td>0</td>\n      <td>0</td>\n    </tr>\n    <tr>\n      <th>1</th>\n      <td>1</td>\n      <td>2.0</td>\n      <td>1</td>\n      <td>11</td>\n      <td>1</td>\n      <td>65</td>\n      <td>27</td>\n      <td>1</td>\n      <td>2.247994</td>\n      <td>1</td>\n      <td>...</td>\n      <td>0</td>\n      <td>0</td>\n      <td>0</td>\n      <td>0</td>\n      <td>0</td>\n      <td>0</td>\n      <td>0</td>\n      <td>3.000000</td>\n      <td>1</td>\n      <td>0</td>\n    </tr>\n    <tr>\n      <th>2</th>\n      <td>1</td>\n      <td>2.0</td>\n      <td>1</td>\n      <td>6</td>\n      <td>1</td>\n      <td>65</td>\n      <td>27</td>\n      <td>2</td>\n      <td>1.000000</td>\n      <td>1</td>\n      <td>...</td>\n      <td>0</td>\n      <td>0</td>\n      <td>0</td>\n      <td>0</td>\n      <td>0</td>\n      <td>1</td>\n      <td>0</td>\n      <td>3.000000</td>\n      <td>1</td>\n      <td>0</td>\n    </tr>\n    <tr>\n      <th>3</th>\n      <td>1</td>\n      <td>1.0</td>\n      <td>1</td>\n      <td>3</td>\n      <td>1</td>\n      <td>3</td>\n      <td>2</td>\n      <td>1</td>\n      <td>2.247994</td>\n      <td>1</td>\n      <td>...</td>\n      <td>0</td>\n      <td>0</td>\n      <td>0</td>\n      <td>0</td>\n      <td>0</td>\n      <td>0</td>\n      <td>0</td>\n      <td>3.079814</td>\n      <td>0</td>\n      <td>0</td>\n    </tr>\n    <tr>\n      <th>4</th>\n      <td>1</td>\n      <td>1.0</td>\n      <td>1</td>\n      <td>5</td>\n      <td>1</td>\n      <td>83</td>\n      <td>3</td>\n      <td>1</td>\n      <td>2.247994</td>\n      <td>1</td>\n      <td>...</td>\n      <td>0</td>\n      <td>0</td>\n      <td>0</td>\n      <td>1</td>\n      <td>0</td>\n      <td>0</td>\n      <td>0</td>\n      <td>3.079814</td>\n      <td>0</td>\n      <td>1</td>\n    </tr>\n  </tbody>\n</table>\n<p>5 rows × 254 columns</p>\n</div>"},"metadata":{}}]},{"metadata":{},"cell_type":"markdown","source":"## SCALE THE FEATURES BETWEEN 0-1\n\nScaling is very important to range the columns into same range. MinMaxScaler ranges the data between 0 to 1."},{"metadata":{"trusted":true},"cell_type":"code","source":"# INITIALISE SCALER\nscaler = MinMaxScaler()\n\n# CREATE DATAFRAME OF THE SCALED DATA\nscaled_train_data = pd.DataFrame(scaler.fit_transform(filtered_data), index=filtered_data.index, columns=filtered_data.columns)\nscaled_train_data.head()","execution_count":27,"outputs":[{"output_type":"execute_result","execution_count":27,"data":{"text/plain":"   month  var8  a.1       a.2  a.3       a.4      var5  b.5       b.6  b.7  \\\n0    0.0   0.5  0.0  0.363636  0.0  0.407407  0.673469  0.0  0.408091  0.0   \n1    0.0   0.5  0.0  0.272727  0.0  0.407407  0.673469  1.0  0.666667  0.0   \n2    0.0   0.5  0.0  0.363636  0.0  0.407407  0.836735  1.0  0.000000  0.0   \n3    0.0   0.5  0.0  0.454545  0.0  0.407407  0.673469  1.0  1.000000  0.0   \n4    0.0   0.5  0.0  0.636364  0.0  0.148148  0.000000  0.0  0.408091  0.0   \n\n   ...  c.277  c.278  c.279  c.280  c.281  c.282  c.283     f.284  Mono  Multi  \n0  ...    0.0    0.0    0.0    1.0    0.0    0.0    0.0  1.000000   1.0    0.0  \n1  ...    0.0    0.0    0.0    0.0    0.0    1.0    0.0  0.509634   1.0    0.0  \n2  ...    0.0    0.0    0.0    0.0    0.0    0.0    0.0  0.500000   0.0    1.0  \n3  ...    0.0    0.0    0.0    1.0    0.0    1.0    0.0  1.000000   1.0    0.0  \n4  ...    0.0    0.0    0.0    0.0    0.0    0.0    0.0  1.000000   0.0    1.0  \n\n[5 rows x 254 columns]","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>month</th>\n      <th>var8</th>\n      <th>a.1</th>\n      <th>a.2</th>\n      <th>a.3</th>\n      <th>a.4</th>\n      <th>var5</th>\n      <th>b.5</th>\n      <th>b.6</th>\n      <th>b.7</th>\n      <th>...</th>\n      <th>c.277</th>\n      <th>c.278</th>\n      <th>c.279</th>\n      <th>c.280</th>\n      <th>c.281</th>\n      <th>c.282</th>\n      <th>c.283</th>\n      <th>f.284</th>\n      <th>Mono</th>\n      <th>Multi</th>\n    </tr>\n  </thead>\n  <tbody>\n    <tr>\n      <th>0</th>\n      <td>0.0</td>\n      <td>0.5</td>\n      <td>0.0</td>\n      <td>0.363636</td>\n      <td>0.0</td>\n      <td>0.407407</td>\n      <td>0.673469</td>\n      <td>0.0</td>\n      <td>0.408091</td>\n      <td>0.0</td>\n      <td>...</td>\n      <td>0.0</td>\n      <td>0.0</td>\n      <td>0.0</td>\n      <td>1.0</td>\n      <td>0.0</td>\n      <td>0.0</td>\n      <td>0.0</td>\n      <td>1.000000</td>\n      <td>1.0</td>\n      <td>0.0</td>\n    </tr>\n    <tr>\n      <th>1</th>\n      <td>0.0</td>\n      <td>0.5</td>\n      <td>0.0</td>\n      <td>0.272727</td>\n      <td>0.0</td>\n      <td>0.407407</td>\n      <td>0.673469</td>\n      <td>1.0</td>\n      <td>0.666667</td>\n      <td>0.0</td>\n      <td>...</td>\n      <td>0.0</td>\n      <td>0.0</td>\n      <td>0.0</td>\n      <td>0.0</td>\n      <td>0.0</td>\n      <td>1.0</td>\n      <td>0.0</td>\n      <td>0.509634</td>\n      <td>1.0</td>\n      <td>0.0</td>\n    </tr>\n    <tr>\n      <th>2</th>\n      <td>0.0</td>\n      <td>0.5</td>\n      <td>0.0</td>\n      <td>0.363636</td>\n      <td>0.0</td>\n      <td>0.407407</td>\n      <td>0.836735</td>\n      <td>1.0</td>\n      <td>0.000000</td>\n      <td>0.0</td>\n      <td>...</td>\n      <td>0.0</td>\n      <td>0.0</td>\n      <td>0.0</td>\n      <td>0.0</td>\n      <td>0.0</td>\n      <td>0.0</td>\n      <td>0.0</td>\n      <td>0.500000</td>\n      <td>0.0</td>\n      <td>1.0</td>\n    </tr>\n    <tr>\n      <th>3</th>\n      <td>0.0</td>\n      <td>0.5</td>\n      <td>0.0</td>\n      <td>0.454545</td>\n      <td>0.0</td>\n      <td>0.407407</td>\n      <td>0.673469</td>\n      <td>1.0</td>\n      <td>1.000000</td>\n      <td>0.0</td>\n      <td>...</td>\n      <td>0.0</td>\n      <td>0.0</td>\n      <td>0.0</td>\n      <td>1.0</td>\n      <td>0.0</td>\n      <td>1.0</td>\n      <td>0.0</td>\n      <td>1.000000</td>\n      <td>1.0</td>\n      <td>0.0</td>\n    </tr>\n    <tr>\n      <th>4</th>\n      <td>0.0</td>\n      <td>0.5</td>\n      <td>0.0</td>\n      <td>0.636364</td>\n      <td>0.0</td>\n      <td>0.148148</td>\n      <td>0.000000</td>\n      <td>0.0</td>\n      <td>0.408091</td>\n      <td>0.0</td>\n      <td>...</td>\n      <td>0.0</td>\n      <td>0.0</td>\n      <td>0.0</td>\n      <td>0.0</td>\n      <td>0.0</td>\n      <td>0.0</td>\n      <td>0.0</td>\n      <td>1.000000</td>\n      <td>0.0</td>\n      <td>1.0</td>\n    </tr>\n  </tbody>\n</table>\n<p>5 rows × 254 columns</p>\n</div>"},"metadata":{}}]},{"metadata":{},"cell_type":"markdown","source":"## SAME FOR THE TEST DATA"},{"metadata":{"trusted":true},"cell_type":"code","source":"# CREATE DATAFRAME OF THE SCALED DATA\nscaled_test_data = pd.DataFrame(scaler.fit_transform(filtered_test_data), index=filtered_test_data.index, columns=filtered_test_data.columns)\nscaled_test_data.head()","execution_count":28,"outputs":[{"output_type":"execute_result","execution_count":28,"data":{"text/plain":"   month  var8  a.1       a.2  a.3       a.4      var5  b.5       b.6  b.7  \\\n0    0.0   0.0  0.0  0.909091  0.0  0.328358  0.142857  1.0  0.000000  0.0   \n1    0.0   0.5  0.0  0.909091  0.0  0.462687  0.530612  0.0  0.415998  0.0   \n2    0.0   0.5  0.0  0.454545  0.0  0.462687  0.530612  1.0  0.000000  0.0   \n3    0.0   0.0  0.0  0.181818  0.0  0.000000  0.020408  0.0  0.415998  0.0   \n4    0.0   0.0  0.0  0.363636  0.0  0.597015  0.040816  0.0  0.415998  0.0   \n\n   ...  c.277  c.278  c.279  c.280  c.281  c.282  c.283     f.284  Mono  Multi  \n0  ...    0.0    0.0    0.0    0.0    0.0    1.0    0.0  0.500000   0.0    0.0  \n1  ...    0.0    0.0    0.0    0.0    0.0    0.0    0.0  0.500000   1.0    0.0  \n2  ...    0.0    0.0    0.0    0.0    0.0    1.0    0.0  0.500000   1.0    0.0  \n3  ...    0.0    0.0    0.0    0.0    0.0    0.0    0.0  0.519953   0.0    0.0  \n4  ...    0.0    0.0    0.0    1.0    0.0    0.0    0.0  0.519953   0.0    1.0  \n\n[5 rows x 254 columns]","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>month</th>\n      <th>var8</th>\n      <th>a.1</th>\n      <th>a.2</th>\n      <th>a.3</th>\n      <th>a.4</th>\n      <th>var5</th>\n      <th>b.5</th>\n      <th>b.6</th>\n      <th>b.7</th>\n      <th>...</th>\n      <th>c.277</th>\n      <th>c.278</th>\n      <th>c.279</th>\n      <th>c.280</th>\n      <th>c.281</th>\n      <th>c.282</th>\n      <th>c.283</th>\n      <th>f.284</th>\n      <th>Mono</th>\n      <th>Multi</th>\n    </tr>\n  </thead>\n  <tbody>\n    <tr>\n      <th>0</th>\n      <td>0.0</td>\n      <td>0.0</td>\n      <td>0.0</td>\n      <td>0.909091</td>\n      <td>0.0</td>\n      <td>0.328358</td>\n      <td>0.142857</td>\n      <td>1.0</td>\n      <td>0.000000</td>\n      <td>0.0</td>\n      <td>...</td>\n      <td>0.0</td>\n      <td>0.0</td>\n      <td>0.0</td>\n      <td>0.0</td>\n      <td>0.0</td>\n      <td>1.0</td>\n      <td>0.0</td>\n      <td>0.500000</td>\n      <td>0.0</td>\n      <td>0.0</td>\n    </tr>\n    <tr>\n      <th>1</th>\n      <td>0.0</td>\n      <td>0.5</td>\n      <td>0.0</td>\n      <td>0.909091</td>\n      <td>0.0</td>\n      <td>0.462687</td>\n      <td>0.530612</td>\n      <td>0.0</td>\n      <td>0.415998</td>\n      <td>0.0</td>\n      <td>...</td>\n      <td>0.0</td>\n      <td>0.0</td>\n      <td>0.0</td>\n      <td>0.0</td>\n      <td>0.0</td>\n      <td>0.0</td>\n      <td>0.0</td>\n      <td>0.500000</td>\n      <td>1.0</td>\n      <td>0.0</td>\n    </tr>\n    <tr>\n      <th>2</th>\n      <td>0.0</td>\n      <td>0.5</td>\n      <td>0.0</td>\n      <td>0.454545</td>\n      <td>0.0</td>\n      <td>0.462687</td>\n      <td>0.530612</td>\n      <td>1.0</td>\n      <td>0.000000</td>\n      <td>0.0</td>\n      <td>...</td>\n      <td>0.0</td>\n      <td>0.0</td>\n      <td>0.0</td>\n      <td>0.0</td>\n      <td>0.0</td>\n      <td>1.0</td>\n      <td>0.0</td>\n      <td>0.500000</td>\n      <td>1.0</td>\n      <td>0.0</td>\n    </tr>\n    <tr>\n      <th>3</th>\n      <td>0.0</td>\n      <td>0.0</td>\n      <td>0.0</td>\n      <td>0.181818</td>\n      <td>0.0</td>\n      <td>0.000000</td>\n      <td>0.020408</td>\n      <td>0.0</td>\n      <td>0.415998</td>\n      <td>0.0</td>\n      <td>...</td>\n      <td>0.0</td>\n      <td>0.0</td>\n      <td>0.0</td>\n      <td>0.0</td>\n      <td>0.0</td>\n      <td>0.0</td>\n      <td>0.0</td>\n      <td>0.519953</td>\n      <td>0.0</td>\n      <td>0.0</td>\n    </tr>\n    <tr>\n      <th>4</th>\n      <td>0.0</td>\n      <td>0.0</td>\n      <td>0.0</td>\n      <td>0.363636</td>\n      <td>0.0</td>\n      <td>0.597015</td>\n      <td>0.040816</td>\n      <td>0.0</td>\n      <td>0.415998</td>\n      <td>0.0</td>\n      <td>...</td>\n      <td>0.0</td>\n      <td>0.0</td>\n      <td>0.0</td>\n      <td>1.0</td>\n      <td>0.0</td>\n      <td>0.0</td>\n      <td>0.0</td>\n      <td>0.519953</td>\n      <td>0.0</td>\n      <td>1.0</td>\n    </tr>\n  </tbody>\n</table>\n<p>5 rows × 254 columns</p>\n</div>"},"metadata":{}}]},{"metadata":{},"cell_type":"markdown","source":"## CHECK LABEL OR TARGET COLUMN\n\nCheck the distribution of the 6 classes among the dataset"},{"metadata":{"trusted":true},"cell_type":"code","source":"# PLOT DISTRIBUTION OF THE TARGET CLASSES\nplt.figure(figsize=(15,6))\nsns.countplot(label.pov6, palette='Set2')\nplt.show()\n\n# SUMMARIZE DISTRIBUTION\ncounter = Counter(label.pov6)\nfor k,v in counter.items():\n    per = v / len(label) * 100\n    print('Class=%d, n=%d (%.3f%%)' % (k, v, per))","execution_count":29,"outputs":[{"output_type":"display_data","data":{"text/plain":"<Figure size 1080x432 with 1 Axes>","image/png":"iVBORw0KGgoAAAANSUhEUgAAA4gAAAFzCAYAAAB8eic9AAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4yLjEsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy+j8jraAAAcPElEQVR4nO3dcbBm9Vkf8O+T3QRJFANhQdylLhoaC9Q2YYcS00kzQcuqMTA26GaasI2021KaxmonBZ1p2nF2Gq0agxVmGEhYkjQEiSnohFRKjKkWIUsSJUAwW0lhhWQ3JkbUEQs+/eP+dvpmuWzuXfbec/fez2fmnfe8zzm/c58z887Ad3/n/N7q7gAAAMBzpm4AAACAlUFABAAAIImACAAAwCAgAgAAkERABAAAYBAQAQAASJKsn7qB5XbiiSf25s2bp24DAABgEvfcc8+XunvDfPvWXEDcvHlzdu/ePXUbAAAAk6iq//NM+9xiCgAAQBIBEQAAgEFABAAAIImACAAAwCAgAgAAkERABAAAYBAQAQAASCIgAgAAMAiIAAAAJBEQAQAAGAREAAAAkgiIAAAADAIiAAAASZL1UzdwtPiJ226YugWehZ//vounbgEAAFa8JZtBrKp3VdW+qvrMPPv+bVV1VZ04U7uiqvZU1YNVdf5M/eyqunfsu7KqatSPqaoPjPpdVbV5qa4FAABgLVjKW0yvT7L14GJVnZrke5M8PFM7I8m2JGeOMVdV1bqx++okO5KcPl4HznlJkq9094uTvCPJzyzJVQAAAKwRSxYQu/vjSb48z653JHlrkp6pXZDkxu5+orsfSrInyTlVdUqS47r7zu7uJDckuXBmzK6xfXOS8w7MLgIAALB4y7pITVW9NskfdffvHbRrY5JHZj7vHbWNY/vg+teM6e4nk3w1yYue4e/uqKrdVbV7//79z/o6AAAAVqNlC4hV9fwkP5Xk38+3e55aH6J+qDFPL3Zf091bunvLhg0bFtIuAADAmrOcM4jfkeS0JL9XVZ9PsinJJ6vqWzI3M3jqzLGbkjw66pvmqWd2TFWtT/LNmf+WVgAAABZg2QJid9/b3Sd19+bu3py5gPey7v5CkluTbBsrk56WucVo7u7ux5I8XlXnjucLL05yyzjlrUm2j+3XJfnoeE4RAACAw7CUP3Px/iR3JnlJVe2tqkue6djuvi/JTUnuT/KRJJd191Nj96VJrs3cwjX/O8lto35dkhdV1Z4kP57k8iW5EAAAgDVi/VKduLtf/3X2bz7o884kO+c5bneSs+ap/2WSi55dlwAAABywrKuYAgAAsHIJiAAAACQREAEAABgERAAAAJIIiAAAAAwCIgAAAEkERAAAAAYBEQAAgCQCIgAAAIOACAAAQBIBEQAAgEFABAAAIImACAAAwCAgAgAAkERABAAAYBAQAQAASCIgAgAAMAiIAAAAJBEQAQAAGAREAAAAkgiIAAAADAIiAAAASQREAAAABgERAACAJAIiAAAAg4AIAABAEgERAACAQUAEAAAgiYAIAADAICACAACQREAEAABgEBABAABIIiACAAAwCIgAAAAkERABAAAYliwgVtW7qmpfVX1mpvafq+qzVfX7VfWhqnrhzL4rqmpPVT1YVefP1M+uqnvHviurqkb9mKr6wKjfVVWbl+paAAAA1oKlnEG8PsnWg2q3Jzmru78ryR8kuSJJquqMJNuSnDnGXFVV68aYq5PsSHL6eB045yVJvtLdL07yjiQ/s2RXAgAAsAYsWUDs7o8n+fJBtd/o7ifHx99NsmlsX5Dkxu5+orsfSrInyTlVdUqS47r7zu7uJDckuXBmzK6xfXOS8w7MLgIAALB4Uz6D+KNJbhvbG5M8MrNv76htHNsH179mzAidX03yovn+UFXtqKrdVbV7//79R+wCAAAAVpNJAmJV/VSSJ5O870BpnsP6EPVDjXl6sfua7t7S3Vs2bNiw2HYBAADWhGUPiFW1PclrkvzjcdtoMjczeOrMYZuSPDrqm+apf82Yqlqf5Jtz0C2tAAAALNyyBsSq2prk3yV5bXf/xcyuW5NsGyuTnpa5xWju7u7HkjxeVeeO5wsvTnLLzJjtY/t1ST46EzgBAABYpPVLdeKqen+SVyU5sar2Jnlb5lYtPSbJ7WM9md/t7n/R3fdV1U1J7s/craeXdfdT41SXZm5F1GMz98zigecWr0vynqrak7mZw21LdS0AAABrwZIFxO5+/Tzl6w5x/M4kO+ep705y1jz1v0xy0bPpEQAAgP9vylVMAQAAWEEERAAAAJIIiAAAAAwCIgAAAEkERAAAAAYBEQAAgCQCIgAAAIOACAAAQBIBEQAAgEFABAAAIImACAAAwCAgAgAAkERABAAAYBAQAQAASCIgAgAAMAiIAAAAJBEQAQAAGAREAAAAkgiIAAAADAIiAAAASQREAAAABgERAACAJAIiAAAAg4AIAABAEgERAACAQUAEAAAgiYAIAADAICACAACQREAEAABgEBABAABIIiACAAAwCIgAAAAkERABAAAYBEQAAACSCIgAAAAMSxYQq+pdVbWvqj4zUzuhqm6vqs+N9+Nn9l1RVXuq6sGqOn+mfnZV3Tv2XVlVNerHVNUHRv2uqtq8VNcCAACwFizlDOL1SbYeVLs8yR3dfXqSO8bnVNUZSbYlOXOMuaqq1o0xVyfZkeT08TpwzkuSfKW7X5zkHUl+ZsmuBAAAYA1YsoDY3R9P8uWDyhck2TW2dyW5cKZ+Y3c/0d0PJdmT5JyqOiXJcd19Z3d3khsOGnPgXDcnOe/A7CIAAACLt9zPIJ7c3Y8lyXg/adQ3Jnlk5ri9o7ZxbB9c/5ox3f1kkq8medF8f7SqdlTV7qravX///iN0KQAAAKvLSlmkZr6Zvz5E/VBjnl7svqa7t3T3lg0bNhxmiwAAAKvbcgfEL47bRjPe94363iSnzhy3Kcmjo75pnvrXjKmq9Um+OU+/pRUAAIAFWu6AeGuS7WN7e5JbZurbxsqkp2VuMZq7x22oj1fVueP5wosPGnPgXK9L8tHxnCIAAACHYf1Snbiq3p/kVUlOrKq9Sd6W5O1JbqqqS5I8nOSiJOnu+6rqpiT3J3kyyWXd/dQ41aWZWxH12CS3jVeSXJfkPVW1J3Mzh9uW6loAAADWgiULiN39+mfYdd4zHL8zyc556ruTnDVP/S8zAiYAAADP3kpZpAYAAICJCYgAAAAkERABAAAYBEQAAACSCIgAAAAMAiIAAABJBEQAAAAGAREAAIAkAiIAAACDgAgAAEASAREAAIBBQAQAACCJgAgAAMAgIAIAAJBEQAQAAGAQEAEAAEgiIAIAADAIiAAAACQREAEAABgERAAAAJIIiAAAAAwCIgAAAEkERAAAAAYBEQAAgCQCIgAAAIOACAAAQBIBEQAAgEFABAAAIImACAAAwCAgAgAAkERABAAAYBAQAQAASCIgAgAAMAiIAAAAJJkoIFbVv6mq+6rqM1X1/qr6hqo6oapur6rPjffjZ46/oqr2VNWDVXX+TP3sqrp37LuyqmqK6wEAAFgNlj0gVtXGJP86yZbuPivJuiTbklye5I7uPj3JHeNzquqMsf/MJFuTXFVV68bprk6yI8np47V1GS8FAABgVZnqFtP1SY6tqvVJnp/k0SQXJNk19u9KcuHYviDJjd39RHc/lGRPknOq6pQkx3X3nd3dSW6YGQMAAMAiLXtA7O4/SvJzSR5O8liSr3b3byQ5ubsfG8c8luSkMWRjkkdmTrF31DaO7YPrAAAAHIYFBcSqumMhtQWe6/jMzQqeluRbk7ygqt5wqCHz1PoQ9fn+5o6q2l1Vu/fv37/YlgEAANaEQwbEA4vHJDmxqo4fC8mcUFWbMxfuDsf3JHmou/d39/9N8qtJvjvJF8dtoxnv+8bxe5OcOjN+U+ZuSd07tg+uP013X9PdW7p7y4YNGw6zbQAAgNXt680g/vMk9yT5zvF+4HVLkl8+zL/5cJJzq+r5Y9XR85I8kOTWJNvHMdvH38iob6uqY6rqtMwtRnP3uA318ao6d5zn4pkxAAAALNL6Q+3s7ncmeWdVvbm7f+lI/MHuvquqbk7yySRPJvlUkmuSfGOSm6rqksyFyIvG8fdV1U1J7h/HX9bdT43TXZrk+iTHJrltvAAAADgMhwyIB3T3L1XVdyfZPDumu284nD/a3W9L8raDyk9kbjZxvuN3Jtk5T313krMOpwcAAAC+1oICYlW9J8l3JPl0kgOzdwd+WgIAAIBVYEEBMcmWJGeM3xsEAABgFVro7yB+Jsm3LGUjAAAATGuhM4gnJrm/qu7O3LOCSZLufu2SdAUAAMCyW2hA/A9L2QQAAADTW+gqpr+11I0AAAAwrYWuYvp45lYtTZLnJXlukj/v7uOWqjEAAACW10JnEL9p9nNVXZjknCXpCAAAgEksdBXTr9Hd/y3Jq49wLwAAAExoobeY/tDMx+dk7ncR/SYiAADAKrLQVUx/cGb7ySSfT3LBEe8GAACAySz0GcQ3LXUjAAAATGtBzyBW1aaq+lBV7auqL1bVB6tq01I3BwAAwPJZ6CI1705ya5JvTbIxya+NGgAAAKvEQgPihu5+d3c/OV7XJ9mwhH0BAACwzBYaEL9UVW+oqnXj9YYkf7yUjQEAALC8FhoQfzTJDyf5QpLHkrwuiYVrAAAAVpGF/szFTyfZ3t1fSZKqOiHJz2UuOAIAALAKLHQG8bsOhMMk6e4vJ3np0rQEAADAFBYaEJ9TVccf+DBmEBc6+wgAAMBRYKEh7+eT/K+qujlJZ+55xJ1L1hUAAADLbkEBsbtvqKrdSV6dpJL8UHffv6SdAQAAsKwWfJvoCIRCIQAAwCq10GcQAQAAWOUERAAAAJIIiAAAAAwCIgAAAEkERAAAAAYBEQAAgCQCIgAAAIOACAAAQBIBEQAAgEFABAAAIImACAAAwCAgAgAAkGSigFhVL6yqm6vqs1X1QFW9vKpOqKrbq+pz4/34meOvqKo9VfVgVZ0/Uz+7qu4d+66sqpriegAAAFaDqWYQ35nkI939nUn+TpIHklye5I7uPj3JHeNzquqMJNuSnJlka5KrqmrdOM/VSXYkOX28ti7nRQAAAKwmyx4Qq+q4JK9Mcl2SdPdfdfefJLkgya5x2K4kF47tC5Lc2N1PdPdDSfYkOaeqTklyXHff2d2d5IaZMQAAACzSFDOI355kf5J3V9WnquraqnpBkpO7+7EkGe8njeM3JnlkZvzeUds4tg+uAwAAcBimCIjrk7wsydXd/dIkf55xO+kzmO+5wj5E/eknqNpRVburavf+/fsX2y8AAMCaMEVA3Jtkb3ffNT7fnLnA+MVx22jG+76Z40+dGb8pyaOjvmme+tN09zXdvaW7t2zYsOGIXQgAAMBqsuwBsbu/kOSRqnrJKJ2X5P4ktybZPmrbk9wytm9Nsq2qjqmq0zK3GM3d4zbUx6vq3LF66cUzYwAAAFik9RP93TcneV9VPS/JHyZ5U+bC6k1VdUmSh5NclCTdfV9V3ZS5EPlkksu6+6lxnkuTXJ/k2CS3jRcAAACHYZKA2N2fTrJlnl3nPcPxO5PsnKe+O8lZR7Y7AACAtWmq30EEAABghREQAQAASCIgAgAAMAiIAAAAJBEQAQAAGAREAAAAkgiIAAAADAIiAAAASQREAAAABgERAACAJAIiAAAAg4AIAABAEgERAACAQUAEAAAgiYAIAADAICACAACQREAEAABgEBABAABIIiACAAAwCIgAAAAkERABAAAYBEQAAACSCIgAAAAMAiIAAABJBEQAAAAGAREAAIAkAiIAAACDgAgAAEASAREAAIBBQAQAACCJgAgAAMAgIAIAAJBEQAQAAGAQEAEAAEgiIAIAADBMFhCral1Vfaqqfn18PqGqbq+qz43342eOvaKq9lTVg1V1/kz97Kq6d+y7sqpqimsBAABYDaacQXxLkgdmPl+e5I7uPj3JHeNzquqMJNuSnJlka5KrqmrdGHN1kh1JTh+vrcvTOgAAwOozSUCsqk1JfiDJtTPlC5LsGtu7klw4U7+xu5/o7oeS7ElyTlWdkuS47r6zuzvJDTNjAAAAWKSpZhB/Mclbk/z1TO3k7n4sScb7SaO+MckjM8ftHbWNY/vg+tNU1Y6q2l1Vu/fv339krgAAAGCVWfaAWFWvSbKvu+9Z6JB5an2I+tOL3dd095bu3rJhw4YF/lkAAIC1Zf0Ef/MVSV5bVd+f5BuSHFdV703yxao6pbsfG7eP7hvH701y6sz4TUkeHfVN89QBAAA4DMs+g9jdV3T3pu7enLnFZz7a3W9IcmuS7eOw7UluGdu3JtlWVcdU1WmZW4zm7nEb6uNVde5YvfTimTEAAAAs0hQziM/k7UluqqpLkjyc5KIk6e77quqmJPcneTLJZd391BhzaZLrkxyb5LbxAgAA4DBMGhC7+2NJPja2/zjJec9w3M4kO+ep705y1tJ1CAAAsHZM+TuIAAAArCACIgAAAEkERAAAAAYBEQAAgCQCIgAAAIOACAAAQBIBEQAAgEFABAAAIImACAAAwCAgAgAAkERABAAAYFg/dQOwGu27+q1Tt8BhOunSn526BQCAyZhBBAAAIImACAAAwCAgAgAAkERABAAAYBAQAQAASCIgAgAAMAiIAAAAJBEQAQAAGAREAAAAkgiIAAAADAIiAAAASQREAAAABgERAACAJAIiAAAAg4AIAABAEgERAACAQUAEAAAgiYAIAADAICACAACQREAEAABgEBABAABIIiACAAAwCIgAAAAkmSAgVtWpVfWbVfVAVd1XVW8Z9ROq6vaq+tx4P35mzBVVtaeqHqyq82fqZ1fVvWPflVVVy309AAAAq8UUM4hPJvmJ7v5bSc5NcllVnZHk8iR3dPfpSe4YnzP2bUtyZpKtSa6qqnXjXFcn2ZHk9PHaupwXAgAAsJose0Ds7se6+5Nj+/EkDyTZmOSCJLvGYbuSXDi2L0hyY3c/0d0PJdmT5JyqOiXJcd19Z3d3khtmxgAAALBIkz6DWFWbk7w0yV1JTu7ux5K5EJnkpHHYxiSPzAzbO2obx/bB9fn+zo6q2l1Vu/fv338kLwEAAGDVmCwgVtU3Jvlgkh/r7j891KHz1PoQ9acXu6/p7i3dvWXDhg2LbxYAAGANmCQgVtVzMxcO39fdvzrKXxy3jWa87xv1vUlOnRm+Kcmjo75pnjoAAACHYYpVTCvJdUke6O5fmNl1a5LtY3t7kltm6tuq6piqOi1zi9HcPW5Dfbyqzh3nvHhmDAAAAIu0foK/+Yokb0xyb1V9etR+Msnbk9xUVZckeTjJRUnS3fdV1U1J7s/cCqiXdfdTY9ylSa5PcmyS28YLAACAw7DsAbG7fzvzPz+YJOc9w5idSXbOU9+d5Kwj1x0AAMDaNekqpgAAAKwcAiIAAABJBEQAAAAGAREAAIAkAiIAAACDgAgAAEASAREAAIBBQAQAACCJgAgAAMAgIAIAAJBEQAQAAGAQEAEAAEgiIAIAADAIiAAAACQREAEAABgERAAAAJIIiAAAAAwCIgAAAEkERAAAAAYBEQAAgCQCIgAAAIOACAAAQBIBEQAAgEFABAAAIImACAAAwCAgAgAAkERABAAAYBAQAQAASCIgAgAAMAiIAAAAJEnWT90AwFp21Xt/e+oWeBb+5Rv+/tQtAMARZQYRAACAJGYQAQBgTXv07u+fugUO07ee8+Ejfk4ziAAAACRZBTOIVbU1yTuTrEtybXe/feKWAGBJ7Ln6t6ZugcP04kv/wdQtACzIUR0Qq2pdkl9O8r1J9ib5RFXd2t33T9sZAMB03nPXjqlb4Fl449+7ZuoWWMOO9ltMz0myp7v/sLv/KsmNSS6YuCcAAICj0tEeEDcmeWTm895RAwAAYJGqu6fu4bBV1UVJzu/ufzo+vzHJOd395oOO25HkwL0WL0ny4LI2enQ4McmXpm6Co4LvCovh+8JC+a6wGL4vLJTvyvy+rbs3zLfjqH4GMXMzhqfOfN6U5NGDD+rua5K4mfsQqmp3d2+Zug9WPt8VFsP3hYXyXWExfF9YKN+VxTvabzH9RJLTq+q0qnpekm1Jbp24JwAAgKPSUT2D2N1PVtW/SvLfM/czF+/q7vsmbgsAAOCodFQHxCTp7g8n+fDUfawCbsFloXxXWAzfFxbKd4XF8H1hoXxXFumoXqQGAACAI+dofwYRAACAI0RAXOOq6l1Vta+qPjN1L6xsVXVqVf1mVT1QVfdV1Vum7omVqaq+oarurqrfG9+V/zh1T6x8VbWuqj5VVb8+dS+sXFX1+aq6t6o+XVW7p+6Hla2qXlhVN1fVZ8f/v7x86p6OBm4xXeOq6pVJ/izJDd191tT9sHJV1SlJTunuT1bVNyW5J8mF3X3/xK2xwlRVJXlBd/9ZVT03yW8neUt3/+7ErbGCVdWPJ9mS5Ljufs3U/bAyVdXnk2zpbr9rx9dVVbuS/M/uvnb84sHzu/tPpu5rpTODuMZ198eTfHnqPlj5uvux7v7k2H48yQNJNk7bFStRz/mz8fG54+VfI3lGVbUpyQ8kuXbqXoDVoaqOS/LKJNclSXf/lXC4MAIisGhVtTnJS5PcNW0nrFTjdsFPJ9mX5Pbu9l3hUH4xyVuT/PXUjbDidZLfqKp7qmrH1M2won17kv1J3j1uX7+2ql4wdVNHAwERWJSq+sYkH0zyY939p1P3w8rU3U91999NsinJOVXlFnbmVVWvSbKvu++ZuheOCq/o7pcl+b4kl41HZWA+65O8LMnV3f3SJH+e5PJpWzo6CIjAgo3nyT6Y5H3d/atT98PKN27n+ViSrRO3wsr1iiSvHc+W3Zjk1VX13mlbYqXq7kfH+74kH0pyzrQdsYLtTbJ35g6WmzMXGPk6BERgQcbCI9cleaC7f2Hqfli5qmpDVb1wbB+b5HuSfHbarlipuvuK7t7U3ZuTbEvy0e5+w8RtsQJV1QvGImkZtwr+wyRWYWde3f2FJI9U1UtG6bwkFtZbgPVTN8C0qur9SV6V5MSq2pvkbd193bRdsUK9Iskbk9w7ni1Lkp/s7g9P2BMr0ylJdlXVusz9Q+RN3e2nC4Bn6+QkH5r798qsT/Jfu/sj07bECvfmJO8bK5j+YZI3TdzPUcHPXAAAAJDELaYAAAAMAiIAAABJBEQAAAAGAREAAIAkAiIAAACDgAgAK0RVPa+qrqmqP6iqz1bVP5q6JwDWFr+DCAArx08l2dfdf7OqnpPkhKkbAmBtMYMIAM9SVW0eM367qur3q+rmqnp+VZ1XVZ+qqnur6l1VdUxVfV9V3TQz9lVV9Wvj448m+U9J0t1/3d1fmuJ6AFi7BEQAODJekuSa7v6uJH+a5MeTXJ/kR7r7b2furp1Lk9ye5NyqesEY9yNJPlBVLxyff7qqPllVv1JVJy/rFQCw5gmIAHBkPNLdvzO235vkvCQPdfcfjNquJK/s7ieTfCTJD1bV+iQ/kOSWzAXITUl+p7tfluTOJD+3nBcAAAIiABwZvYhjP5Dkh5O8OsknuvvxJH+c5C+SfGgc8ytJXnZEOwSAr0NABIAj429U1cvH9uuT/I8km6vqxaP2xiS/NbY/lrnw988yFxbT3Z3k15K8ahxzXpL7l7xrAJhRc/89AgAOV1VtTvLhJB9P8t1JPpe5QPjyzN0muj7JJ5Jc2t1PjDH/Jck/SXJSd//FqH1bkvckeWGS/Une1N0PL+OlALDGCYgA8CyNgPjr3X3WxK0AwLPiFlMAAACSmEEEAABgMIMIAABAEgERAACAQUAEAAAgiYAIAADAICACAACQREAEAABg+H8GLohFQHei9QAAAABJRU5ErkJggg==\n"},"metadata":{"needs_background":"light"}},{"output_type":"stream","text":"Class=1, n=14197 (77.246%)\nClass=2, n=3246 (17.661%)\nClass=4, n=183 (0.996%)\nClass=6, n=285 (1.551%)\nClass=3, n=335 (1.823%)\nClass=5, n=133 (0.724%)\n","name":"stdout"}]},{"metadata":{},"cell_type":"markdown","source":"As we can see, it is highly unbalanced data with 77 % of the data in class 1 and rest of the data falls in other classes. So we will use resampling technique. \n\nResampling techniques include undersampling and oversampling. **Undersampling** means reducing the majority classes to the minority class and **Oversampling** means increasing the minority classes to the majority class.\n\nWe will not use Undersampling because minority class has 133 datapoints and reducing to such a less number will reduce our datapoints which is not good for the classification. Hence we will go for Oversampling."},{"metadata":{},"cell_type":"markdown","source":"## BALANCE THE LABEL DATA BY OVERSAMPLING & UNDERSAMPLING "},{"metadata":{},"cell_type":"markdown","source":"## SMOTE "},{"metadata":{"trusted":true},"cell_type":"code","source":"# PERFORM OVERSAMPLING USING SMOTE ( SYNTHETIC MINORITY OVERSAMPLING TECHNIQUE )\noversample = SMOTE()\nX_over, y_over = oversample.fit_resample(scaled_train_data, label)\n\n# PLOT THE NEW DATA\nplt.figure(figsize=(10,6))\nsns.countplot(y_over.pov6, palette='Wistia')\nplt.show()\n\n# SUMMARIZE DISTRIBUTION\ncounter = Counter(y_over.pov6)\nfor k,v in counter.items():\n    per = v / len(y_over) * 100\n    print('Class=%d, n=%d (%.3f%%)' % (k, v, per))","execution_count":30,"outputs":[{"output_type":"display_data","data":{"text/plain":"<Figure size 720x432 with 1 Axes>","image/png":"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\n"},"metadata":{"needs_background":"light"}},{"output_type":"stream","text":"Class=1, n=14197 (16.667%)\nClass=2, n=14197 (16.667%)\nClass=4, n=14197 (16.667%)\nClass=6, n=14197 (16.667%)\nClass=3, n=14197 (16.667%)\nClass=5, n=14197 (16.667%)\n","name":"stdout"}]},{"metadata":{},"cell_type":"markdown","source":"## NearMiss 1"},{"metadata":{"trusted":true},"cell_type":"code","source":"undersample = NearMiss(version=1, n_neighbors=3)\n# transform the dataset\nX_under, y_under = undersample.fit_resample(scaled_train_data, label)\n\n# PLOT THE NEW DATA\nplt.figure(figsize=(10,6))\nsns.countplot(y_under.pov6, palette='autumn')\nplt.show()\n\n# SUMMARIZE DISTRIBUTION\ncounter = Counter(y_under.pov6)\nfor k,v in counter.items():\n    per = v / len(y_under) * 100\n    print('Class=%d, n=%d (%.3f%%)' % (k, v, per))","execution_count":31,"outputs":[{"output_type":"display_data","data":{"text/plain":"<Figure size 720x432 with 1 Axes>","image/png":"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\n"},"metadata":{"needs_background":"light"}},{"output_type":"stream","text":"Class=1, n=133 (16.667%)\nClass=2, n=133 (16.667%)\nClass=3, n=133 (16.667%)\nClass=4, n=133 (16.667%)\nClass=5, n=133 (16.667%)\nClass=6, n=133 (16.667%)\n","name":"stdout"}]},{"metadata":{},"cell_type":"markdown","source":"Now the data is balanced."},{"metadata":{},"cell_type":"markdown","source":"## CLASSIFICATION "},{"metadata":{},"cell_type":"markdown","source":"## TEST MULTIPLE CLASSIFICATION ALGORITHMS ON UNDERSAMPLED DATA\n\nWe will test multiple classification algorithms on undersampled data and choose the algorithm with high F beta score. F beta score is a generalized form of F1 score. F1 score has precision and recall equal weightage. In Fbeta score, you can give weightage to precision and recall using parameter beta. We will give Recall higher preference. \n\nThen the best algorithm will be fit on both UNDERSAMPLED and OVERSAMPLED data and we will compare which resampling technique gives the best results."},{"metadata":{"trusted":true},"cell_type":"code","source":"# CREATE PIPELINE OF ALL THE POSSIBLE CLASSIFIERS\nest =[]\n\nest.append(('SVC', Pipeline([('SVC', SVC(gamma='scale', class_weight='balanced'))])))\nest.append(('GradientBoosting', Pipeline([('GradientBoosting',GradientBoostingClassifier())])))\nest.append(('AdaBoost', Pipeline([ ('AdaBoost', AdaBoostClassifier())])))\nest.append(('ExtraTree', Pipeline([('ExtraTrees', ExtraTreesClassifier())])))\nest.append(('RandomForest', Pipeline([('RandomForest', RandomForestClassifier())]))) \nest.append(('Bagging', Pipeline([('Bagging', BaggingClassifier())])))\nest.append(('KNeighbors', Pipeline([('KNeighbors', KNeighborsClassifier())])))\nest.append(('DecisionTree', Pipeline([('DecisionTree', DecisionTreeClassifier())])))\n# est.append(('XGB', Pipeline([('XGB', XGBClassifier())])))","execution_count":32,"outputs":[]},{"metadata":{},"cell_type":"markdown","source":"We will use Repeated Stratified K-Fold model validation technique to train the model. It repeats Stratified K-Fold n times with different randomization in each repetition. Stratified K-Fold cross-validation object is a variation of KFold that returns stratified folds. The folds are made by preserving the percentage of samples for each class. \n\nIt will make sure that model is overfitted."},{"metadata":{},"cell_type":"markdown","source":"## Train the UNDERSAMPLED data with different classifiers"},{"metadata":{"trusted":true},"cell_type":"code","source":"import warnings\nwarnings.filterwarnings(action='ignore')\n\nseed = 4\nsplits = 5\n\n# CALCULATE F1 SCORE WITH MORE WEIGHT ON RECALL\nf_score = make_scorer(fbeta_score, beta=2, average='macro')\nmodels_score =[]\n\nfor i in est:\n    cv = RepeatedStratifiedKFold(n_splits=5, n_repeats=3, random_state=1)\n    results = cross_val_score(i[1], X_under, y_under, cv=cv, scoring=f_score)\n    models_score.append({i[0] : '{} +/- {}'.format(results.mean(), results.std())})","execution_count":33,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"print(\"F1 SCORES OF DIFFERENT CLASSIFIERS\")\nmodels_score","execution_count":34,"outputs":[{"output_type":"stream","text":"F1 SCORES OF DIFFERENT CLASSIFIERS\n","name":"stdout"},{"output_type":"execute_result","execution_count":34,"data":{"text/plain":"[{'SVC': '0.7364519447217158 +/- 0.027315537428438463'},\n {'GradientBoosting': '0.8918568151906339 +/- 0.025585579208257863'},\n {'AdaBoost': '0.5386930183605194 +/- 0.1456766826124554'},\n {'ExtraTree': '0.8102332447952961 +/- 0.03502519994553447'},\n {'RandomForest': '0.8190086627722997 +/- 0.033911821956679994'},\n {'Bagging': '0.8484276274728455 +/- 0.03199955801350012'},\n {'KNeighbors': '0.6133847410554039 +/- 0.040020123924571334'},\n {'DecisionTree': '0.7910144280473919 +/- 0.0341990977627551'}]"},"metadata":{}}]},{"metadata":{},"cell_type":"markdown","source":"As we can see GradientBoosting (89%) , Bagging (84%) has great Fbeta scores. We will choose the one with the highest Fbeta score. "},{"metadata":{},"cell_type":"markdown","source":"## FIT THE BEST MODEL ON THE UNDERSAMPLED DATA"},{"metadata":{"trusted":true},"cell_type":"code","source":"model = GradientBoostingClassifier()\nmodel.fit(X_under, y_under.pov6)\n# PREDICT ON THE TEST DATA\npredict_under = model.predict(scaled_test_data)","execution_count":35,"outputs":[]},{"metadata":{},"cell_type":"markdown","source":"## FIT THE BEST MODEL ON THE OVERSAMPLED DATA"},{"metadata":{"trusted":true},"cell_type":"code","source":"model_ = GradientBoostingClassifier()\nmodel_.fit(X_over, y_over.pov6)\n# PREDICT ON THE TEST DATA\npredict_over = model_.predict(scaled_test_data)","execution_count":36,"outputs":[]},{"metadata":{},"cell_type":"markdown","source":"## UNDERSAMPLED vs OVERSAMPLED PREDICTED DISTRIBUTION COMPARISON"},{"metadata":{"trusted":true},"cell_type":"code","source":"l = [predict_under, predict_over, train_data.pov6]\npal = [\"autumn\",'Wistia','Set2']\ntitles =['UNDERSAMPLED', 'OVERSAMPLED', 'TRAINING']\n\nplt.figure(figsize=(25,6))\nplt.rc('font', size=15)\nfor i in range(3):\n    plt.subplot(1,3, i+1)\n    sns.countplot(l[i], palette=pal[i])\n    plt.title(titles[i])\n    plt.xlabel('CLASSES')\n    plt.ylabel('COUNT')\nplt.show()\n\n# SUMMARIZE DISTRIBUTION\nprint('DISTRIBUTION OF PREDICTED UNDERSAMPLED CLASSES')\ncounter = Counter(predict_under)\nfor k,v in counter.items():\n    per = v / len(predict_under) * 100\n    print('Class=%d, n=%d (%.3f%%)' % (k, v, per))\n    \nprint()\nprint('DISTRIBUTION OF PREDICTED OVERSAMPLED CLASSES')\ncounter = Counter(predict_over)\nfor k,v in counter.items():\n    per = v / len(predict_over) * 100\n    print('Class=%d, n=%d (%.3f%%)' % (k, v, per))","execution_count":37,"outputs":[{"output_type":"display_data","data":{"text/plain":"<Figure size 1800x432 with 3 Axes>","image/png":"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\n"},"metadata":{"needs_background":"light"}},{"output_type":"stream","text":"DISTRIBUTION OF PREDICTED UNDERSAMPLED CLASSES\nClass=1, n=2053 (44.679%)\nClass=2, n=783 (17.040%)\nClass=4, n=165 (3.591%)\nClass=6, n=378 (8.226%)\nClass=5, n=1126 (24.505%)\nClass=3, n=90 (1.959%)\n\nDISTRIBUTION OF PREDICTED OVERSAMPLED CLASSES\nClass=1, n=3479 (75.713%)\nClass=4, n=55 (1.197%)\nClass=2, n=854 (18.585%)\nClass=3, n=80 (1.741%)\nClass=6, n=84 (1.828%)\nClass=5, n=43 (0.936%)\n","name":"stdout"}]},{"metadata":{},"cell_type":"markdown","source":"### As we can see OVERSAMPLED data retains the same distribution of the classes same as the training data. Thus we can conclude that the OVERSAMPLING technique is not good.\n\n### Whereas UNDERSAMPLED data, gives a good result as the classes are not skewed. Let's experiment with some more undersampled techniques."},{"metadata":{},"cell_type":"markdown","source":"## DIFFERENT UNDERSAMPLED TECHNIQUES\n\n### * NearMiss 2\n     NearMiss(version=2, n_neighbors=3)\n### * NearMiss 3\n     NearMiss(version=3, n_neighbors=3)\n### * Condensed Nearest Neighbour\n     CondensedNearestNeighbour(n_neighbors=1)\n### * Tomek Links\n     TomekLinks()\n### * Edited Nearest Neighbours\n     EditedNearestNeighbours(n_neighbors=3)\n\n## COMBINED TECHNIQUES \n\n### * One Sided Selection   (Tomek Links and the Condensed Nearest Neighbor (CNN))\n     OneSidedSelection(n_neighbors=1, n_seeds_S=200)\n### * Neighborhood Cleaning Rule (Condensed Nearest Neighbor & Edited Nearest Neighbors )\n     NeighbourhoodCleaningRule(n_neighbors=3, threshold_cleaning=0.5)"},{"metadata":{},"cell_type":"markdown","source":"## Alert ! Following code takes a lot of time "},{"metadata":{"trusted":true},"cell_type":"code","source":"# NEAR MISS VERSION 2\nundersample = NearMiss(version=2, n_neighbors=3)\n# transform the dataset\nX_near2, y_near2 = undersample.fit_resample(scaled_train_data, label)\n\n# NEAR MISS VERSION 3\nundersample = NearMiss(version=3, n_neighbors=3)\n# transform the dataset\nX_near3, y_near3 = undersample.fit_resample(scaled_train_data, label)\n\n# CONDENSED NEAREST NEIGHBOUR\nundersample = CondensedNearestNeighbour(n_neighbors=1)\n# transform the dataset\nX_cnn, y_cnn = undersample.fit_resample(scaled_train_data, label)\n\n# TOMEK LINKS\nundersample = TomekLinks()\n# transform the dataset\nX_tomek, y_tomek = undersample.fit_resample(scaled_train_data, label)\n\n# EDITED NEAREST NEIGHBOURS\nundersample = EditedNearestNeighbours(n_neighbors=3)\n# transform the dataset\nX_enn, y_enn = undersample.fit_resample(scaled_train_data, label)\n\n# ONE SIDED SELECTION \nundersample = OneSidedSelection(n_neighbors=1, n_seeds_S=200)\n# transform the dataset\nX_oss, y_oss = undersample.fit_resample(scaled_train_data, label)\n\n# NEIGHBOURHOOD CLEANING RULE\nundersample = NeighbourhoodCleaningRule(n_neighbors=3, threshold_cleaning=0.5)\n# transform the dataset\nX_ncr, y_ncr = undersample.fit_resample(scaled_train_data, label)","execution_count":38,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"model = GradientBoostingClassifier()\n\n# FIT ON NEAR MISS VERSION 2\nmodel.fit(X_near2, y_near2)\n# PREDICT ON THE TEST DATA\npredict_near2 = model.predict(scaled_test_data)\n\n# NEAR MISS VERSION 3\nmodel.fit(X_near3, y_near3)\npredict_near3 = model.predict(scaled_test_data)\n\n# CONDENSED NEAREST NEIGHBOUR\nmodel.fit(X_cnn, y_cnn)\npredict_cnn = model.predict(scaled_test_data)\n\n# TOMEK LINKS\nmodel.fit(X_tomek, y_tomek)\npredict_tomek = model.predict(scaled_test_data)\n\n# EDITED NEAREST NEIGHBOURS\nmodel.fit(X_enn, y_enn)\npredict_enn = model.predict(scaled_test_data)\n\n# ONE SIDED SELECTION\nmodel.fit(X_oss, y_oss)\npredict_oss = model.predict(scaled_test_data)\n\n# NEIGHBOURHOOD CLEANING RULE\nmodel.fit(X_ncr, y_ncr)\npredict_ncr = model.predict(scaled_test_data)","execution_count":39,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"pred_list = [predict_near2, predict_near3, predict_cnn, predict_tomek, predict_enn, predict_oss, predict_ncr]\npal = ['autumn','Blues','Reds_r', 'icefire', 'Wistia','Purples', 'Greens']\ntitles =['NEAR_2', 'NEAR_3', 'Condensed Nearest Neighbour', 'TOMEK_LINKS', 'Edited Nearest Neighbour', 'One Sided Selection', 'Neighborhood Cleaning Rule']\n\nplt.figure(figsize=(25,25))\nplt.rc('font', size=15)\nfor i in range(len(pred_list)):\n    plt.subplot(3, 3, i+1)\n    sns.countplot(pred_list[i], palette=pal[i])\n    plt.title(titles[i])\n    plt.xlabel('CLASSES')\n    plt.ylabel('COUNT')\nplt.show()","execution_count":43,"outputs":[{"output_type":"display_data","data":{"text/plain":"<Figure size 1800x1800 with 7 Axes>","image/png":"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\n"},"metadata":{"needs_background":"light"}}]},{"metadata":{},"cell_type":"markdown","source":"### We can see from the prediction plot except Near Miss version 2, no one has given non skewed values"},{"metadata":{},"cell_type":"markdown","source":"## COMPARISON BETWEEN NearMiss 1 vs NearMiss 2"},{"metadata":{"trusted":true},"cell_type":"code","source":"l = [predict_under, predict_near2]\npal = ['Set1','Set2']\ntitles =['NearMiss 1', 'NearMiss 2']\n\nplt.figure(figsize=(25,6))\nplt.rc('font', size=15)\nfor i in range(2):\n    plt.subplot(1,2, i+1)\n    sns.countplot(l[i], palette=pal[i])\n    plt.title(titles[i])\n    plt.xlabel('CLASSES')\n    plt.ylabel('COUNT')\nplt.show()\n\nprint('NearMiss Version 1')\ncounter = Counter(predict_under)\nfor k,v in counter.items():\n    per = v / len(predict_under) * 100\n    print('Class=%d, n=%d (%.3f%%)' % (k, v, per))\n\nprint()\nprint()\nprint('NearMiss Version 2')\ncounter = Counter(predict_near2)\nfor k,v in counter.items():\n    per = v / len(predict_near2) * 100\n    print('Class=%d, n=%d (%.3f%%)' % (k, v, per))","execution_count":46,"outputs":[{"output_type":"display_data","data":{"text/plain":"<Figure size 1800x432 with 2 Axes>","image/png":"iVBORw0KGgoAAAANSUhEUgAABcEAAAGRCAYAAACt7HudAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4yLjEsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy+j8jraAAAgAElEQVR4nOzde9yeVX3v+c+3oQFiCQkQrDpEUNoiUk/NbpVpy2nYFhzK9rTB1vHAdCI6FJ10YyFNxoiloK0iJXsXGevWbT0hWpm4iZHYcqhAIVh1bAil2AgKUnA/kIYEA+E3f1zXAzc393PKc76fz/v1ul95rrV+17rX5V+Lr+teV6oKSZIkSZIkSZL60c9M9wQkSZIkSZIkSZoshuCSJEmSJEmSpL5lCC5JkiRJkiRJ6luG4JIkSZIkSZKkvmUILkmSJEmSJEnqW4bgkiRJkiRJkqS+ZQguSRqTJGuSVJJDp3sukiRJksbPNb6kfmcILkkTIMmx7aKxkvzeEDWV5KtTPbdeOub6/w1T853Buqmc21CSnJjksiS3Jnm0ndux0z0vSZIk9SfX+JMryT5J/o8kVyXZmmRnku8n+VySF033/CT1F0NwSZp470+y73RPYhQeBY5K8u+6O5L8CvCStqbbHwP7Aj+Y3Ok9w+8CZwDzgNun+LslSZI0t7nGn3iHApcDBwB/CZwFfA54NfDtJMdN4Vwk9TlDcEmaWJuA5wLvme6JdEuyb5K9OppuAH4CvL1H+RnAg23N01TV41X1aFVN9e6RPwL2q6pXAJ+Z4u+WJEnS3OUaf3I8ALy8qn6jqj5QVX9ZVX8E/Hrb/6dTOBdJfc4QXJIm1hXAbcAfJjlwNDckWZbkr5M8mOSnSe5I8kddi1mS/GqSTyb5pyQ7kvxbkm8meW2PMT/Z/sxxSZJPJLkfeAT4nzrKdtGEyW9Ksk/HvXsDb2r7dvUY+xnnBSY5IMnFSe5qjyr5SZLbkpzTde9bktyS5KEkj7Q/d/xMkiUj/e9UVT+qqp+OVCdJkiRNMNf4k7DGr6qfVNW3e7RvBr4HHDXc/ZI0FobgkjSxCvhDYH+ancvDSnIy8E3gF4EPA2cDNwHn0/wUsNNrgSNoFuHvBi6g+engl5P8zhBfcQ3NrpUPAOcB27v6/xJY1I7d+T2LgU+MNP8OX6T5+eJ64Pfb+d8CHDtYkOTNwKdofn75f9PspPkM8EvAwWP4LkmSJGkqucafwjV+kp8BngPcvyf3S1Ive41cIkkai6r6RpJrgHcluaSqep6r1+7M+ATw98DxVfV42/WxJN8BPpLk2Kq6tm3/46o6r2uMPwf+AVgFfLbH13yvqt48zFy/m+RbND+XHFyQnwHc1vaN+LxJ9geOB/6iqs4apvR1wL/x9GcFWD3il0iSJEnTyDX+kCZjjf9OmhD8A+MYQ5Kexp3gkjQ5/hCYz/ALtxOBZwP/FViU5KDBD3B1W/PvB4ur6pHBv5MsaH+KuQD4G+BFSRb2+I4/G8VcPwGckOSQJIcAJzC2HSI7gZ8Cv9b588keHm7n+5qMZuUtSZIkzSyu8Z9pQtf4SY6m2T3/XeBPxjueJA0yBJekSVBV/0Cz6+J3k7xkiLIXtf9+gualMJ2fLW3fsweLkxyc5PKOs/8ebGvPbEsW9fiOfxrFdD8LPAa8FXgbzRmB3T/THFJV7aL52eNRwL8k+ccklyY5oav0T2jeNv8V4IEkX0rye0n2G+13SZIkSdPFNf7krvGT/Arw34F7gZOr6tGxjiFJQ/E4FEmaPKuANwAfBE7q0T+4U+Ic4BkvhGndC9Duqvg6zaL6z4FbaXZd7Kb5mePv0OP/2KyqHSNNsqoGknyFZnEc4CtVNTDSfV1jXJbkKuA1wDE0z31Wki9U1eltzZ1JjqTZhXJCW/f/AO9P8ptVdddYvlOSJEmaBq7xJ2GNn+QVNGedPwwcV1U/GstcJWkkhuCSNEmq6l+S/AXw7iTH9Si5s/33karaOMJwLwFeCpxfVe/r7Ejye+OfLZ8ATmv/PnO4wqFU1X3Ax4GPJ5kHfJrmrfQfrqpb25qf0vwM9Gp48qVB/x1YAfyf43oCSZIkaZK5xp/4NX6Sl9ME4P9GE4D3PG9dksbD41AkaXL9MbCNZqdItw3AvwLnJjmguzPJvh0/I9w92NxVcxRPf+v7ntpI8/Ka1cA3xnJje3bhgs62qtpNc44fNG+3pz0Hsdu3OmskSZKkWcA1/gSt8dsAfCPNUTDHVdW/jGWekjRa7gSXpElUVQ8m+VN6vDynqh5J8haa8/PuSPIJ4J9pzv07guZN668FrgVuB/4ReG+7GL0D+EXgHcD3gFeMc55P0Czm98QvAtcl+et2LgM0P+l8J/AvwA1t3deTPAxcD9xD85xvA4pmR8mw2nMXf7u9/J/bf/+3JL/e/n1pVT28h88gSZIkjYpr/IlZ4yd5Ps0O8MU0x8Ec3b4Ys9Nfd748VJL2lCG4JE2+jwDvAp7T3VFVG5L8O+Bc4M3AEpoF5l3tfd9t63YneQ3Nm+DfCjyLZjH6VpqfUI5rgTxO99D81PI44D8AewM/ojkL8IMdZxb+BfAfaRb1BwA/Af4B+P2q+ttRfM8reOZ/aJzR8fdf0ZwhKEmSJE021/iN8azxDwMObP9eM0yNIbikcUtVTfccJEmSJEmSJEmaFJ4JLkmSJEmSJEnqW4bgkiRJkiRJkqS+ZQguSZIkSZIkSepbhuCSJEmSJEmSpL5lCC5JkiRJkiRJ6lt7TfcEZrKDDjqoDj300OmehiRJkibYbbfd9mBVLZnueWjqucaXJEnqT8Ot8Q3Bh3HooYeyadOm6Z6GJEmSJliSH0z3HDQ9XONLkiT1p+HW+B6HIkmSJEmSJEnqW4bgkiRJkiRJkqS+ZQguSZIkSZIkSepbhuCSJEmSJEmSpL5lCC5JkiRJkiRJ6luG4JIkSZIkSZKkvmUILkmSJEmSJEnqW4bgkiRJkiRJkqS+ZQguSZIkSZIkSepbhuCSJEmSJEmSpL5lCC5JkiRJkiRJ6luG4JIkSZIkSZKkvmUILkmSJEmSJEnqW3tN9wT6xaZlvzrdU+gryzbdMt1TkCRJ0hz3B+v/23RPoa98+KS3TPcUJEnSHOVOcEmSJEmSJElS3zIElyRJkiRJkiT1rSkNwZO8Mcn/m+RHSbYnuS3Jm7pqkmRlknuS7ExyfZKX9RjryCTfSLIjyb1Jzk8yb0/GkiRJkiRJkiT1p6neCb4C2A78X8BvA38LfDbJ73fUnAusBj4InNLWb0zy84MFSRYDG4ECTgXOB/4AeH/X9404liRJkiRJkiSpf031izFPqaoHO67/JslzacLxS5PsQxNcX1hVawGS3ARsBc4CVrX3nQnsC7yuqrYB1yRZCKxJ8qGq2jaGsSRJkiRJkiRJfWpKd4J3BeCD/gE4uP37aGAhcEXHPY8A64CTOu45CdjQBuCDPk8TjB8zxrEkSZIkSZIkSX1qJrwY82hgc/v3EcBu4M6umtvbPjrqtnQWVNXdwI6OutGOJUmSJEmSJEnqU9Magic5geZM7//cNi0GtlfV7q7SAWBBkvkddQ/1GHKg7RvLWN1zWp5kU5JNDzzwwNgeSJIkSZIkSZI0o0xbCJ7kUOCzwFVV9cmOrupV3qNvqLrR1AzVR1VdXlXLqmrZkiVLepVIkiRJkiRJkmaJaQnBkxwArAfuBt7c0TUA7JdkXtcti4AdVfVYR92iHkPvz1M7xEc7liRJkiRJkiSpT015CJ5kAfBVYD7wmvZllYO2APOAw7tu6z4DfAtd53onOQR4VkfdaMeSJEmSJEmSJPWpKQ3Bk+wFfBH4BeCkqvrXrpIbgW3AGzvuWQCcQrNzfNB64NVJ9utoOw3YCVw3xrEkSZIkSZIkSX1qryn+vv8CnAy8GzggySs7+v6hqh5NchGwOskAzY7tFTRh/aUdtZcBZwNfTvJB4AXAGuAjVbUNYAxjSZIkSZIkSZL61FSH4P++/feSHn2HAVuBi2iC6vOAA4FNwIlVdf9gYVUNJDkBWAusozkH/GKaILzTiGNJkiRJkiRJkvrXlIbgVXXoKGoKuKD9DFe3GTh+IsaSJEmSJEmSJPWnKX8xpiRJkiRJkiRJU8UQXJIkSZIkSZLUtwzBJUmSJI1LksOTfCzJd5LsTnJtV/+xSWqIz4aOurcNUXNm13hJsjLJPUl2Jrk+ycum6HElSZI0y0z1izElSZIk9Z8XAycDNwPze/R/C3hVV9tS4AvA+h71xwM7O66/39V/LrAaOAfYAqwANiY5qqp+PObZS5Ikqa8ZgkuSJEkar3VVdRVAkiuBgzo7q2obTUD+pCS/ATwBXNFjvFuranuvL0qyD00IfmFVrW3bbgK2AmcBq8b1JJIkSeo7HociSZIkaVyq6ok9uO1NwHVVde8Y7zsaWEhHeF5VjwDrgJP2YB6SJEnqc4bgkiRJkqZUkl8AXg58boiSu5I8nuSOJO/o6jsC2A3c2dV+e9snSZIkPY3HoUiSJEmaam8CHgO+1NV+H81Z37cA89q6y5IsqKqL25rFwPaq2t117wCwIMn8qto1eVOXJEnSbGMILkmSJGmqnQ58var+R2djVW0ANnQ0rU+yN7AqySUdx65UjzEzVF+S5cBygKVLl4537pIkSZplPA5FkiRJ0pRJ8lLgRQx9FEq3K4EDgEPb6wFgvyTzuuoWATuq6rHuAarq8qpaVlXLlixZsmcTlyRJ0qxlCC5JkiRpKp0O7ASuGuN9gzu8t9AclXJ4V/8RbZ8kSZL0NIbgkiRJkqbSacC6qto+yvrXAw8CP2ivbwS2AW8cLEiyADgFWD+B85QkSVKf8ExwSZIkSePShtAnt5fPAxYmeUN7fXVV7WjrXgkcBqwYYpwv0bwU87s0u71Paz9nD54HXlWPJrkIWJ1kgGb39wqaDT6XTsLjSZIkaZYzBJckSZI0XgcDX+xqG7w+DNja/n068DBD79i+AzgDOITmRZebgbdU1ae76i6iCb3PAw4ENgEnVtX9e/4IkiRJ6leG4JIkSZLGpaq20oTWI9W9B3jPMP0rgZWjGKeAC9qPJEmSNCzPBJckSZIkSZIk9S1DcEmSJEmSJElS3zIElyRJkiRJkiT1LUNwSZIkSZIkSVLfMgSXJEmSJEmSJPUtQ3BJkiRJkiRJUt8yBJckSZIkSZIk9S1DcEmSJEmSJElS3zIElyRJkiRJkiT1LUNwSZIkSZIkSVLfmvIQPMnhST6W5DtJdie5tqv/2CQ1xGdDR93bhqg5s2u8JFmZ5J4kO5Ncn+RlU/S4kiRJkiRJkqRptNc0fOeLgZOBm4H5Pfq/Bbyqq20p8AVgfY/644GdHdff7+o/F1gNnANsAVYAG5McVVU/HvPsJUmSJEmSJEmzxnSE4Ouq6iqAJFcCB3V2VtU2moD8SUl+A3gCuKLHeLdW1fZeX5RkH5oQ/MKqWtu23QRsBc4CVo3rSSRJkiRJkiRJM9qUH4dSVU/swW1vAq6rqnvHeN/RwEI6wvOqegRYB5y0B/OQJEmSJEmSJM0iM/7FmEl+AXg58LkhSu5K8niSO5K8o6vvCGA3cGdX++1tnyRJkiRJkiSpj03HcShj9SbgMeBLXe330Zz1fQswr627LMmCqrq4rVkMbK+q3V33DgALksyvql2dHUmWA8sBli5dOqEPIkmSJEmSJEmaWrMhBD8d+HpV/Y/OxqraAGzoaFqfZG9gVZJLOo5dqR5jZqi+qrocuBxg2bJlve6VJEmSJEmSJM0SM/o4lCQvBV7E0EehdLsSOAA4tL0eAPZLMq+rbhGwo6oem4h5SpIkSZIkSZJmphkdgtPsAt8JXDXG+wZ3cG+hOSrl8K7+I9o+SZIkSZIkSVIfm+kh+GnAuqraPsr61wMPAj9or28EtgFvHCxIsgA4BVg/gfOUJEmSJEmSJM1AU34meBtCn9xePg9YmOQN7fXVVbWjrXslcBiwYohxvkTzUszv0uz2Pq39nD14HnhVPZrkImB1kgGa3d8raML/Syfh8SRJkiRJkiRJM8h0vBjzYOCLXW2D14cBW9u/TwceZugd23cAZwCH0LzocjPwlqr6dFfdRTSh93nAgcAm4MSqun/PH0GSJEmSJEmSNBtMeQheVVtpQuuR6t4DvGeY/pXAylGMU8AF7UeSJEmSJEmSNIfM9DPBJUmSJEmSJEnaY4bgkiRJkiRJkqS+ZQguSZIkSZIkSepbhuCSJEmSJEmSpL5lCC5JkiRJkiRJ6luG4JIkSZLGJcnhST6W5DtJdie5tkfN1iTV9flxj7ojk3wjyY4k9yY5P8m8rpokWZnkniQ7k1yf5GWT+IiSJEmaxfaa7glIkiRJmvVeDJwM3AzMH6bus8ClHde7OjuTLAY2ApuBU4EXAh+m2byzqqP0XGA1cA6wBVgBbExyVFU9I1iXJEnS3GYILkmSJGm81lXVVQBJrgQOGqLuvqq6eZhxzgT2BV5XVduAa5IsBNYk+VBVbUuyD00IfmFVrW2/8yZgK3AWTw/LJUmSJI9DkSRJkjQ+VfXEBA11ErChDcAHfZ4mGD+mvT4aWAhc0fH9jwDr2vslSZKkpzEElyRJkjRVzkiyK8nDSa5M8vyu/iNojjd5UlXdDexo+wZrdgN3dt17e0eNJEmS9CSPQ5EkSZI0Fa6iOTP8h8CLgPcBNyT55ap6uK1ZDDzU496Btm+wZntV7e5RsyDJ/KrahSRJktQyBJckSZI06arq3R2XNyS5Efg28Hbgo52lPW5PV/tQNT37kiwHlgMsXbp0DLOWJElSP/A4FEmSJElTrqq+B9wBvKKjeQBY1KN8f57aIT4A7JdkXlfNImBHVT3W47sur6plVbVsyZIl45+8JEmSZhVDcEmSJEnTqXPn9ha6zvVOcgjwLJ46K3wLMA84vGucZ5wnLkmSJIEhuCRJkqRpkOQo4JeA2zqa1wOvTrJfR9tpwE7guvb6RmAb8MaOsRYAp7T3S5IkSU/jmeCSJEmSxqUNoU9uL58HLEzyhvb6auA44M3AV4F7aXZtrwLuBj7ZMdRlwNnAl5N8EHgBsAb4SFVtA6iqR5NcBKxOMkCz+3sFzQafSyfpESVJkjSLGYJLkiRJGq+DgS92tQ1eHwbc09Z8lObs7p8AXwNWDobbAFU1kOQEYC2wjuYc8ItpgvBOF9GE3ucBBwKbgBOr6v6JeyRJkiT1C0NwSZIkSeNSVVuBjFB2wijH2gwcP0JNARe0H0mSJGlYhuCSJEmSpFnpX//ivdM9hb5z8Ds/NN1TkCRpwvliTEmSJEmSJElS3zIElyRJkiRJkiT1LUNwSZIkSZIkSVLfMgSXJEmSJEmSJPUtQ3BJkiRJkiRJUt+a8hA8yeFJPpbkO0l2J7m2R83WJNX1+XGPuiOTfCPJjiT3Jjk/ybyumiRZmeSeJDuTXJ/kZZP4iJIkSZIkSZKkGWKvafjOFwMnAzcD84ep+yxwacf1rs7OJIuBjcBm4FTghcCHaYL9VR2l5wKrgXOALcAKYGOSo6rqGcG6JEmSJEmSJKl/TEcIvq6qrgJIciVw0BB191XVzcOMcyawL/C6qtoGXJNkIbAmyYeqaluSfWhC8Auram37nTcBW4GzeHpYLkmSJEmSJEnqM1N+HEpVPTFBQ50EbGgD8EGfpwnGj2mvjwYWAld0fP8jwLr2fkmSJEmSJElSH5vJL8Y8I8muJA8nuTLJ87v6j6A53uRJVXU3sKPtG6zZDdzZde/tHTWSJEmSJEmSpD41HcehjMZVNGeG/xB4EfA+4IYkv1xVD7c1i4GHetw70PYN1myvqt09ahYkmV9Vu5AkSZIkSZIk9aUZGYJX1bs7Lm9IciPwbeDtwEc7S3vcnq72oWp69iVZDiwHWLp06RhmLUmSJEmSJEmaaWbycShPqqrvAXcAr+hoHgAW9Sjfn6d2iA8A+yWZ11WzCNhRVY/1+K7Lq2pZVS1bsmTJ+CcvSZIkSZIkSZo2syIE79C5c3sLXed6JzkEeBZPnRW+BZgHHN41zjPOE5ckSZIkSZIk9Z9ZEYInOQr4JeC2jub1wKuT7NfRdhqwE7iuvb4R2Aa8sWOsBcAp7f2SJEmSJEmSpD425WeCtyH0ye3l84CFSd7QXl8NHAe8GfgqcC/Nru1VwN3AJzuGugw4G/hykg8CLwDWAB+pqm0AVfVokouA1UkGaHZ/r6AJ/y+dpEeUJEmSJEmSJM0Q0/FizIOBL3a1DV4fBtzT1nyU5uzunwBfA1YOhtsAVTWQ5ARgLbCO5hzwi2mC8E4X0YTe5wEHApuAE6vq/ol7JEmSJEmSJEnSTDTlIXhVbQUyQtkJoxxrM3D8CDUFXNB+JEmSJEmSJElzyKw4E1ySJEmSJEmSpD1hCC5JkiRJkiRJ6luG4JIkSZIkSZKkvmUILkmSJEmSJEnqW4bgkiRJkiRJkqS+ZQguSZIkSZIkSepbhuCSJEmSJEmSpL5lCC5JkiRJkiRJ6luG4JIkSZIkSZKkvmUILkmSJGlckhye5GNJvpNkd5Jru/qfk+RP2/7tSe5J8qkkz+2qe1uS6vE5s6suSVa24+xMcn2Sl03Bo0qSJGkW2mu6JyBJkiRp1nsxcDJwMzC/R/+vAK8FPg78PfBsYA1wY5Kjqmp7V/3xwM6O6+939Z8LrAbOAbYAK4CN7Vg/Ht+jSJIkqd8YgkuSJEkar3VVdRVAkiuBg7r6/w44oqoeH2xI8i3gDuD1wKe66m/tEYwP3rcPTQh+YVWtbdtuArYCZwGrxv00kiRJ6isehyJJkiRpXKrqiRH6H+oMwNu2fwJ2AAeP8euOBhYCV3SM9QiwDjhpjGNJkiRpDjAElyRJkjTlkrwEWABs7tF9V5LHk9yR5B1dfUcAu4E7u9pvb/skSZKkp/E4FEmSJElTKsnPAJfQBNlf7+i6j+as71uAecCbgMuSLKiqi9uaxcD2qtrdNewAsCDJ/KraNakPIEmSpFnFEFySJEnSVLsQeBVwTFU9NthYVRuADR1165PsDaxKcknHsSvVY8wM1ZdkObAcYOnSpRMwfUmSJM0mHociSZIkacokeRdwDvDWqvr7UdxyJXAAcGh7PQDsl2ReV90iYEdnqD6oqi6vqmVVtWzJkiV7PnlJkiTNSobgkiRJkqZEktcDlwLvraovjPH2wR3eW2iOSjm8q/+Itk+SJEl6GkNwSZIkSZMuybHAZ4C1VfVnY7j19cCDwA/a6xuBbcAbO8ZeAJwCrJ+QyUqSJKmveCa4JEmSpHFpQ+iT28vnAQuTvKG9vhp4PvAVmp3aX0jyyo7bH6iqu9pxvkTzUszv0uz2Pq39nD14HnhVPZrkImB1koF2zBU0G3wunbynlCRJ0mxlCC5JkiRpvA4GvtjVNnh9GPBrwP7AS4FvdtV9Cnhb+/cdwBnAITQvutwMvKWqPt11z0U0ofd5wIHAJuDEqrp/vA8iSZKk/mMILkmSJGlcqmorTWg9lE+2n5HGWQmsHEVdARe0H0mSJGlYngkuSZIkSZIkSepbhuCSJEmSJEmSpL5lCC5JkiRJkiRJ6ltTHoInOTzJx5J8J8nuJNd29T8nyZ+2/duT3JPkU0me21X3tiTV43NmV12SrGzH2Znk+iQvm4JHlSRJkiRJkiRNs+l4MeaLgZOBm4H5Pfp/BXgt8HHg74FnA2uAG5McVVXbu+qPB3Z2XH+/q/9cYDVwDrAFWAFsbMf68fgeRZIkSZIkSZI0k01HCL6uqq4CSHIlcFBX/98BR1TV44MNSb4F3AG8HvhUV/2tPYLxwfv2oQnBL6yqtW3bTcBW4Cxg1bifRpIkSZIkSZI0Y035cShV9cQI/Q91BuBt2z8BO4CDx/h1RwMLgSs6xnoEWAecNMaxJEmSJEmSJEmzzKx4MWaSlwALgM09uu9K8niSO5K8o6vvCGA3cGdX++1tnyRJkiRJkiSpj03HcShjkuRngEtoguyvd3TdR3PW9y3APOBNwGVJFlTVxW3NYmB7Ve3uGnYAWJBkflXtmtQHkCRJkiRJkiRNmxkfggMXAq8CjqmqxwYbq2oDsKGjbn2SvYFVSS7pOHaleoyZofqSLAeWAyxdunQCpi9JkiRJkiRJmi7DHoeS5G+STNuxIUneBZwDvLWq/n4Ut1wJHAAc2l4PAPslmddVtwjY0RmqD6qqy6tqWVUtW7JkyZ5PXpIkSZqBpnuNL0mSJE21kc4EP5bmxZJTLsnrgUuB91bVF8Z4++AO7y00R6Uc3tV/RNsnSZIkzTXHMk1rfEmSJGk6zMgXYyY5FvgMsLaq/mwMt74eeBD4QXt9I7ANeGPH2AuAU4D1EzJZSZIkSZIkSdKMNZozwXudqb3H2hD65PbyecDCJG9or68Gng98hWan9heSvLLj9geq6q52nC/RvBTzuzS7vU9rP2cPngdeVY8muQhYnWSgHXMFTfh/6UQ+lyRJkjSLTOgaX5IkSZrJRhOCfyXJT0czWFW9YBRlBwNf7GobvD4M+DVgf+ClwDe76j4FvK39+w7gDOAQmhddbgbeUlWf7rrnIprQ+zzgQGATcGJV3T+KuUqSJEn9aKLX+JIkSdKMNZoQ/G+B+ybqC6tqK01oPZRPtp+RxlkJrBxFXQEXtB9JkiRJE7zGlyRJkmay0YTgf15Vt0z6TCRJkiRNFdf4kiRJmjNm5IsxJUmSJEmSJEmaCIbgkiRJkiRJkqS+NVII/n7gh1MxEUmSJElTwjW+JEmS5pRhzwSvqvdP1UQkSZIkTT7X+JIkSZprhg3BkzwB1CjHqqoazYs2JUmSJE0T1/iSJEmaa0Za0J7N8AvkAK8FjpuwGUmSJEmaTK7xJUmSNKeMdBzK2l7tSQKcBvwR8GJgPfDHEz47SZIkSRPKNb4kSZLmmpFejPk0SeYleTuwBfgr4A7gFVX1mqq6aTImKEmSJGnyuMaXJElSvxvV+X5J5gO/B7wXeB7weeA/VNXtkzg3SZIkSZPENb4kSZLmipFejLkAeCfwB8ABwH8DLqqq70/B3CRJkiRNMNf4kiRJmmtG2gm+FTgQuBb4EHAPsE+SI3sVV9XmiZycJEnaM1vXHDbdU+g7h675l+megjRRtuIaX5IkSXPISCH4Qe2/xwHHDlMXmjfMz5uAOUmSJEmaPBO+xk9yOODc8psAACAASURBVHAO8ErgKOCGqjq2qybAeTS70A8CbgXOrqpvd9UdCVwKvAp4CPg48P6q2j3WsSRJkiQYOQQ/bkpmIUmSJGmqTMYa/8XAycDNwPwhas4FVtOE5VuAFcDGJEdV1Y8BkiwGNgKbgVOBFwIfBn4GWDWWsSRJkqRBw4bgVXXdVE1EkiRJ0uSbpDX+uqq6CiDJlTy125y2bR+a4PrCqlrbtt1EczTLWTwVcJ8J7Au8rqq2AdckWQisSfKhqto2hrEkSZIkoNlRIUmSJEl7rKqeGKHkaGAhcEXHPY8A64CTOupOAja0Afigz9ME48eMcSxJkiQJGCEET/JEkt1DfHYl+WGSzyZ5yVRNWJIkSdKem6Y1/hHAbuDOrvbb277Oui2dBVV1N7Cjo260Y0mSJEnAyGeCn03zMpxe5gHPAf5X4OYkv1lVmyZycpIkSZIm3HSs8RcD2ztfbtkaABYkmV9Vu9q6h3rcP9D2jWUsSZIkCRj5TPC1Iw2QZCVwDfAB/PmhJEmSNKNN4xq/V/CeHn1D1Y2mpmdfkuXAcoClS5eOOFFJkiT1l3GfCV5VBVwGvGr805EkSZI03SZhjT8A7JdkXlf7ImBHVT3WUbeox/3789QO8dGO9aSquryqllXVsiVLluzxQ0iSJGl2mqgXYz4M7D1BY0mSJEmafhO5xt9Cc9TK4V3t3WeAb6HrXO8khwDP6qgb7ViSJEkSMHEh+DHAXRM0liRJkqTpN5Fr/BuBbcAbBxuSLABOAdZ31K0HXp1kv46204CdwHVjHEuSJEkCRjgTvF1MDmUe8PPAqcB/Av5wAuclSZIkaRJMxhq/HfPk9vJ5wMIkb2ivr66qHUkuAlYnGaDZsb2CZlPOpR1DXUbz4s4vJ/kg8AJgDfCRqtoGUFWPjnIsSZIkCRghBAe2M/Sb4wftAi6uqksmZkqSJEmSJtFkrPEPBr7Y1TZ4fRiwFbiIJqg+DzgQ2AScWFX3D95QVQNJTgDWAutozgG/mCYI7zTiWJIkSdKgkULwMxh6gfw48ABwS1U9NESNJEmSpJllwtf4VbUVyAg1BVzQfoar2wwcPxFjSZIkSTBCCF5Vn5zoL0xyOHAO8ErgKOCGqjq2qyY0uzreCRwE3AqcXVXf7qo7kuYnj6+i2SXyceD9VbV7rGNJkiRJc8FkrPElSZKkmWykneBPSrIU+HWaM/4AfkQTYN8zxu98Mc15gTcD84eoORdYTROWD57xtzHJUVX143Y+i4GNwGaaMwtfCHyY5meRq8YyliRJkjQXTeAaX5IkSZqxRgzBk/w88F+A36YJmDs9keQrwFljCJTXVdVV7dhX0uzO7vy+fWiC6wuram3bdhPNOYJn8VTAfSawL/C69iU51yRZCKxJ8qGq2jaGsSRJkqQ5YxLW+JIkSdKM1b3gfZokBwJ/B/wmcD7wMmAxsKj9+/3AMcD1SQ4YzRdW1RMjlBwNLASu6LjnEZoX45zUUXcSsGHwLfGtz9ME48eMcSxJkiRpTpiMNb4kSZI0kw0bgtMcI/KzwEuq6vyq+m5VPVxV29q/PwC8hOZYk4naVX0EsBu4s6v99ravs25LZ0FV3Q3s6Kgb7ViSJEnSXDEda3xJkiRp2owUgp8K/HFV3TtUQVXdB/wJ8NoJmtNiYHvnyy1bA8CCJPM76nq9sX6g7RvLWE9KsjzJpiSbHnjggT1+CEmSJGmGmo41viRJkjRtRgrBn0PXbush3N7WTpTq0ZYefUPVjaamZ19VXV5Vy6pq2ZIlS0YzV0mSJGk2ma41viRJkjQtRgrBHwQOHcU4h7a1E2EA2C/JvK72RcCOqnqso25Rj/v356kd4qMdS5IkSZorpmONL0mSJE2bkULwDcB7kzxrqIIkPwe8F1g/QXPaAswDDu9q7z4DfAtd53onOQR4VkfdaMeSJEmS5orpWONLkiRJ02akEHwN8GzgtiS/m2S/wY4kP5fkd4BbgYNp3iI/EW4EtgFv7PiuBcApPH0Rvh54deecgNOAncB1YxxLkiRJmivWMPVrfEmSJGna7DVcZ1Xdk+Q44K+ATwOVZPCokUU0Z2t/Gzi+qn44mi9sQ+iT28vnAQuTvKG9vrqqdiS5CFidZIBmx/YKmsD+0o6hLgPOBr6c5IPAC2gW9B+pqm3t/B8d5ViSJEnSnDAZa3xJkiRpJhs2BAeoqn8EXp7kGOA3aIJrgB8B11fV9WP8zoOBL3a1DV4fBmwFLqIJqs8DDgQ2ASdW1f0d8xpIcgKwFlhHcw74xTRBeKcRx5IkSZLmkklY40uSJEkz1rAheJIDgcuBy6tqA08dM9JZ82pgOfDOqvrXkb6wqrbS7C4ZrqaAC9rPcHWbgeMnYixJkiRpLpiMNb4kSZI0k410Jvh7aI4Z+fowNV+n2cH9BxM1KUmSJEmTxjW+JEmS5pSRQvD/CFzW7qbuqe37GHDqRE5MkiRJ0qRwjS9JkqQ5ZaQQ/PnA5lGMcztw6LhnI0mSJGmyucaXJEnSnDJSCL4TWDiKcX6urZUkSZI0s7nGlyRJ0pwyUgj+LeC3RzHOqW2tJEmSpJnNNb4kSZLmlJFC8P8M/O9J3jpUQZK3AG8H1k7kxCRJkiRNCtf4kiRJmlP2Gq6zqr6c5BLgvyY5C/gacDdQwFLg1cAy4OKq+uvJnqwkSZKk8XGNL0mSpLlm2BAcoKr+IMm1wHuA/wTs3Xb9FPgmcGpVfXXSZihJkiRpQrnGlyRJ0lwyYggOUFXrgHVJ9gIObJt/UlWPT9rMJEmSJE0a1/iSJEmaK0YVgg9qF8T3T9JcJEmSJE0x1/iSJEnqdyO9GFOSJEmSJEmSpFnLEFySJEmSJEmS1LcMwSVJkiRJkiRJfcsQXJIkSZIkSZLUtwzBJUmSJEmSJEl9yxBckiRJ0qRLcm2SGuLzqrZma4++H/cY68gk30iyI8m9Sc5PMm/qn0qSJEmzwV7TPQFJkiRJc8K7gIVdbecDLwdu7Wj7LHBpx/WuzhuSLAY2ApuBU4EXAh+m2eCzamKnLEmSpH5gCC5JkiRp0lXV5s7rJPOBZcAXqurxjq77qurmYYY6E9gXeF1VbQOuSbIQWJPkQ22bJEmS9CSPQ5EkSZI0HX4LWAx8boz3nQRs6Aq7P08TjB8zQXOTJElSHzEElyRJkjQdTgd+BNzQ1X5Gkl1JHk5yZZLnd/UfAWzpbKiqu4EdbZ8kSZL0NB6HIkmSJGlKJVkAnAJcXlXV0XUVcDPwQ+BFwPuAG5L8clU93NYsBh7qMexA2ydJkiQ9jSG4JEmSpKl2CvBzdB2FUlXv7ri8IcmNwLeBtwMf7SztMWaGaCfJcmA5wNKlS/d81pIkSZqVPA5FkiRJ0lQ7Hfjnqto0XFFVfQ+4A3hFR/MAsKhH+f703iFOVV1eVcuqatmSJUv2cMqSJEmarQzBJUmSJE2ZJPvTvNxyLC/E7NzhvYWus7+THAI8i66zwiVJkiQwBJckSZI0tV4L7M0oQvAkRwG/BNzW0bweeHWS/TraTgN2AtdN4DwlSZLUJ2bkmeBJrgWOGaL76Kq6KclWoPtN8fdX1c93jXUkcCnwKpqfR34ceH9V7Z7QSWvG+63VX5juKfSdr33gtOmegiRJmn1OB75TVbd3NiZ5DfBm4KvAvTS7vVcBdwOf7Ci9DDgb+HKSDwIvANYAH6mqbZM9eUmSJM0+MzIEB94FLOxqOx94OXBrR9tnaQLuQbs6b0iyGNgIbAZOBV4IfJhmB/yqiZ2yJEmSpOEkOQg4AVjdo/se4GCaF2AuAn4CfA1Y2RluV9VAkhOAtcA6mo0uF9ME4ZIkSdIzzMgQvKo2d14nmQ8sA75QVY93dN1XVTcPM9SZwL7A69qF8zVJFgJrknzInSKSJEnS1KmqB4GfHaLvuzQB+WjG2QwcP4FTkyRJUh+bLWeC/xawmLG9PAeaF+5s6Aq7P08TjA913IokSZIkSZIkqU/MlhD8dOBHwA1d7Wck2ZXk4SRXJuk+I/wIut4QX1V3AzvoeqO8JEmSJEmSJKn/zMjjUDolWQCcAlxeVdXRdRVwM/BD4EXA+4AbkvxyVT3c1iymOSOw20Db1+v7lgPLAZYuXTohzyBJkiRJkiRJmh6zYSf4KcDP0XUUSlW9u6o+V1U3VNXlwKuB5wJv77q/eKYM0U5VXV5Vy6pq2ZIlS8Y/e0mSJEmSJEnStJkNIfjpwD9X1abhiqrqe8AdwCs6mgdo3izfbX967xCXJEmSJEmSJPWRGR2CJ9mf5uWWY3khZucO7y10nf2d5BDgWXSdFS5JkiRJkiRJ6j8z/Uzw1wJ7M4oQPMlRwC8BH+toXg+ck2S/qvq3tu00YCdw3QTPVZIkSZIkSdIcdu8tJ0/3FPrOc3/16nGPMaN3gtMchfKdqrq9szHJa5J8LsnvJjkuyTuBDcDdwCc7Si8Dfgp8Ocn/0r70cg3wkaraNiVPIEmSJEmSJEmaNjN2J3iSg4ATgNU9uu8BDgY+SnPm90+ArwErO8PtqhpIcgKwFlhHcw74xTRBuCRJkiRJkiSpz83YELyqHgR+doi+79IE5KMZZzNw/AROTZIkSZIkSZI0S8z041AkSZIkSZIkSdpjhuCSJEmSJEmSpL5lCC5JkiRJkiRJ6luG4JIkSZIkSZKkvmUILkmSJEmSJEnqW4bgkiRJkiRJkqS+ZQguSZIkSZIkSepbhuCSJEmSJEmSpL5lCC5JkiRJkiRJ6luG4JIkSZIkSZKkvmUILkmSJEmSJEnqW4bgkiRJkiRJkqS+ZQguSZIkSZIkSepbhuCSJEmSJEmSpL5lCC5JkiRJkiRJ6luG4JIkSZIkSZKkvmUILkmSJEmSJEnqW4bgkiRJkiRJkqS+ZQguSZIkadIleVuS6vE5s6MmSVYmuSfJziTXJ3lZj7GOTPKNJDuS3Jvk/CTzpvaJJEmSNFvsNd0TkCRJkjSnHA/s7Lj+fsff5wKrgXOALcAKYGOSo6rqxwBJFgMbgc3AqcALgQ/TbPBZNemzlyRJ0qxjCC5JkiRpKt1aVdu7G5PsQxOCX1hVa9u2m4CtwFk8FXCfCewLvK6qtgHXJFkIrEnyobZNkiRJepLHoUiSJEmaCY4GFgJXDDZU1SPAOuCkjrqTgA1dYffnaYLxY6ZgnpIkSZplDMElSZIkTaW7kjye5I4k7+hoPwLYDdzZVX9729dZt6WzoKruBnZ01UmSJEmAx6FIkiRJmhr30Zz3fQswD3gTcFmSBVV1MbAY2F5Vu7vuGwAWJJlfVbvauod6jD/Q9kmSJElPYwguSZIkadJV1QZgQ0fT+iR7A6uSXDJY1uPW9Ogbqq5XO0mWA8sBli5dOpZpS5IkqQ/MyONQkrwtSfX4nNlRkyQrk9yTZGeS65O8rMdYRyb5RpIdSe5Ncn6SeVP7RJIkSZJ6uJL/v717D7a9LO8D/n3keMNwB8fEQTBqQ9BOO5YmeKlWISpaQgMiYr1VU0KFwEjjJFBMCY3GS4BQiSLRRkOHwYi2iFMkopXRGi+o0TYIiSISL6jHHkREAfHpH791dLHYB/aBfc76rbU/n5k1Z6/39653PWfe2Xs/5znven7J7kn2zXCSe6cVcvVdk9zS3bdPnm+ajM3aJSufEE93n9fdB3T3AXvttdeaBA4AwOIY+0nwpyf54dTza6e+/r0MH6d8VYaegCclubyqHtfdNyRJVe2W5PIkVyU5LMmjkpyRofh/agAAgDHoDDn9DkkeneSaqWuzPcCvzkzv76raO8lDZuYBAECS8RfBP93dN88OVtWDMhTB/6i7z5mM/XWS65Icn58VuI/NcJf4wyd3j/9gVe2c5LSqesPMHeUBAIDt64gkG5N8NUPP8JuSHJnkD5OkqnZMcmiS86Zec2mSV1XVTt39/cnYURkOz1yxneIGAGCBjLIdyio8McnOSf5y80B3/yDJJUkOmZp3SJLLZordF2YojD91O8QJAAAkqar3VNXvVtUhVfWvqur8DMXr07v7J939oySvS3JKVR1XVQcleXeGf7O8aWqpc5PcmuS9VXXwpN/3aUnOdMgFAICVjP0k+Jerao8kX86Q1L51Mr5fkjuS/P3M/C9mSKQzNe/D0xO6+/qqumVy7ZJtEjUAADDrmiQvS7J3hptYXpXkxd19/tSc12Uoep+cZI8kVyb5te7+1uYJ3b1pUiA/J0M+f2OSszIUwgEA4C7GWgT/ZoZ+35/K0Bfw6CTnVtWO3X1Wkt2S3Nzdd8y8blOSHavqAd1922TeSjfH2TS5dhfuHA8AAGuvu09Jcso9zOkkr5k87m7eVRnuHwQAAPdolEXw7r4syWVTQ5dW1QOTnFpVZ2+etsJLa4VrW5q30ni6+7xMeg4ecMABK84BAAAAAGAxLFJP8IuS7J5k3wwnuXeqqh1m5uya5Jbuvn3yfNNkbNYuWfmEOAAAAAAAS2SRiuCbdZKrM7RJefTMtf0m1za7ejL2U1W1d5KHzMwDAAAAAGAJLVIR/IgkG5N8NcnHk9yU5MjNF6tqxySHJrl06jWXJnlmVe00NXZUkh8muWJbBwwAAAAAwHyNsid4Vb0nw00xv5DhxPdRk8cJ3f2TJD+qqtcleXVVbcpwqvukDEX9N00tdW6SE5K8t6pen+QXM9w1/szuvmk7/XUAAAAAAJiTURbBk1yT5GVJ9s5wE8urkry4u8+fmvO6DEXvk5PskeTKJL/W3d/aPKG7N1XVQUnOSXJJhj7gZ2UohAMAAAAAsORGWQTv7lOSnHIPczrJayaPu5t3VZKnr110AAAAAAAsikXqCQ4AAAAAAFtFERwAAAAAgKWlCA4AAAAAwNJSBAcAAAAAYGkpggMAAAAAsLQUwQEAAAAAWFqK4AAAAAAALC1FcAAAAAAAlpYiOAAAAAAAS0sRHAAAAACApaUIDgAAAADA0tow7wAAANarb37z0HmHsHR+/ucvmXcIAADAyDgJDgAAAADA0lIEBwAAAABgaSmCAwAAAACwtBTBAQAAAABYWorgAAAAAAAsLUVwAAAAAACWliI4AAAAAABLSxEcAAAAAIClpQgOAAAAAMDSUgQHAAC2uao6sqreV1Vfr6qbq+ozVXX0zJzrqqpnHjessNb+VfWhqrqlqr5RVadX1Q7b728DAMAi2TDvAAAAgHXhpCRfSfLKJBuTPDvJBVW1Z3e/aWreBUmmn982vUhV7Zbk8iRXJTksyaOSnJHhgM+p2yx6AAAWliI4AACwPRza3Runnn+4qn4hQ3F8uuj9ze7+xN2sc2ySByc5vLtvSvLBqto5yWlV9YbJGAAA/JR2KAAAwDY3UwDf7HNJHrqVSx2S5LKZYveFGQrjT72X4QEAsMRGWQTXLxAAANaFJ2ZoazLtZVV1W1V9r6ouqqp9Zq7vl+Tq6YHuvj7JLZNrAABwJ2Nth6JfIAAALLGqOihDjv6yqeGLk3wiydeS/HKS/5Tko1X1j7v7e5M5uyW5cYUlN02uAQDAnYy1CK5fIAAALKmq2jfDgZaLu/sdm8e7+8SpaR+tqo8n+Zsk/zbJn0xd65WW3cJ4quqYJMckySMe8Yj7EDkAAItolEXwu+kXeNhWLrWlfoGvz9Av8JJ7FyEAAHBvVNXuSS5Ncn2SF97d3O7+v1V1TZLHTw1vSrLrCtN3yconxNPd5yU5L0kOOOCAFQvlADB253/ymHmHsHRe9KvnzTsEtpNR9gTfAv0CAQBggVXVjknen+QBSZ7T3T9Y5UunC9dXZyaXr6q9kzwkM7k/AAAkC1IEn+oX+KdTwxcnOS7JQUleleQJGT4yucvUnK3uF1hVx1TVlVV15Xe+8521CB8AANa9qtqQ5N1JHpPkkO7+9ipe87gkv5TkM1PDlyZ5ZlXtNDV2VJIfJrli7SIGAGBZjLIdyrTt3S/QRyUBAGCbeHOGG96fmGT3qjpw6trnkhycoT3K+5N8I8Np71MztE15x9Tcc5OckOS9VfX6JL+Y5LQkZ7rnDwAAKxl1EXwe/QIBAIBt4hmTP89e4dojk/xDkodmONCya5LvJvlAklOmi9vdvWnySdFzMtzj58YkZ2UohAMAwF2MtgiuXyAAACyP7t53FdMOWuVaVyV5+n0KCIA7+dJbdJRaS4/+90+ddwjAlFH2BNcvEAAAAACAtTDWk+D6BQIAAAAAcJ+NtQiuXyAAAAAAAPfZKIvg+gUCAAAAALAWRtkTHAAAAAAA1oIiOAAAAAAAS0sRHAAAAACApaUIDgAAAADA0lIEBwAAAABgaW2YdwAALJ6zX/DWeYewVE684LfmHQIAAAAsLSfBAQAAAABYWorgAAAAAAAsLUVwAAAAAACWlp7gAAAAwDbz5v/2sXmHsHRe8cInzzsEgIXiJDgAAAAAAEtLERwAAAAAgKWlCA4AAAAAwNJSBAcAAAAAYGkpggMAAAAAsLQUwQEAAAAAWFob5h0AwLQXnP/8eYewdC540YXzDgEAAABgbpwEBwAAAABgaSmCAwAAAACwtBTBAQAAAABYWorgAAAAAAAsLUVwAAAAAACWliI4AAAAAABLSxEcAAAAAICltS6K4FW1f1V9qKpuqapvVNXpVbXDvOMCAADuHTk+AACrtWHeAWxrVbVbksuTXJXksCSPSnJGhv8AOHWOoQEAAPeCHB8AgK2x9EXwJMcmeXCSw7v7piQfrKqdk5xWVW+YjAEAAItDjg8AwKqth3YohyS5bCYRvjBD0vzU+YQEAADcB3J8AABWbT0UwfdLcvX0QHdfn+SWyTUAAGCxyPEBAFi19VAE3y3JjSuMb5pcAwAAFoscHwCAVavunncM21RV3Z7kd7r77Jnxryd5R3f/x5nxY5IcM3n6S0mu2S6Bbj97Jtk47yC4R/Zp/OzRYrBPi8E+jd8y7tE+3b3XvIPg3pHj38Uyfo8uI/s0fvZoMdinxWCfxm8Z92iLOf56uDHmpiS7rjC+S1Y4PdLd5yU5b1sHNS9VdWV3HzDvOLh79mn87NFisE+LwT6Nnz1ihOT4U3yPLgb7NH72aDHYp8Vgn8Zvve3RemiHcnVm+gJW1d5JHpKZPoIAAMBCkOMDALBq66EIfmmSZ1bVTlNjRyX5YZIr5hMSAABwH8jxAQBYtfVQBD83ya1J3ltVB0/6AZ6W5Mzuvmmukc3H0n4MdMnYp/GzR4vBPi0G+zR+9oixkePfme/RxWCfxs8eLQb7tBjs0/itqz1a+htjJklV7Z/knCRPyNAj8G1JTuvuO+YaGAAAcK/I8QEAWK11UQQHAAAAAGB9Wg/tUNa9qnp0Vb21qj5fVXdU1UfmHRN3VlVHVtX7qurrVXVzVX2mqo6ed1zcWVU9t6o+XlXfraofVdU1VXVqVT1g3rGxsqp6+OR7qqvq5+YdD4OqeulkT2Yfx847Nu6sqjZU1e9V1d9X1a1V9bWqOmvecQFy/EUgx18McvzFI8cfJzn+4livOf6GeQfAdvHYJM9O8okkfpGP00lJvpLklUk2ZtivC6pqz+5+01wjY9oeSf5Xkjdm+Nj1r2ToP/qwJMfPLyzuxhuT3JzkIfMOhBU9PcNN7Da7dl6BsEV/nuSgJH+Q5OokeyfZf64RAZvJ8cdPjr8Y5PiLR44/bnL88VuXOb52KOtAVd2vu38y+fqiJHt297+cb1RMmyTCG2fGLkjyhO5+5JzCYhWq6jVJjkuyW/uBOipV9S+SXJzktRkS5Z26++b5RkUynBLJkHjZkxGrqmcluSTJP+nuq+YdD3Bncvzxk+MvLjn+eMnxx0uOvxjWc46vHco6sDk5Zrxmk+OJzyV56PaOha323Th9NTpVtUOSNyU5PcPJK2DrvSzJh9dbcgyLQo4/fnL8hSbHHyE5PqyJdZvjK4LDeD0xybr7obQIqmqHqtqxqp6c5IQkb3FCZHSOTfKgJH8670C4W1+uqh9Pem/+1ryD4S5+NcnfVdU5VXVTVd1SVe+tql+Yd2AAC0yOP1Jy/IUgx18McvxxW7c5vp7gMEJVdVCSwzL8Dx3j84MkD5x8/RdJXjXHWJhRVXsk+c9JXtjdt1fVvEPirr6Z5NVJPpVkhyRHJzm3qnbs7qW/IcsCeViSlyb5fJLnJ9kpyRuS/PeqOlBhAGDryPFHT44/YnL8hSDHXwzrNsfXE3yd0S9w/Kpq3ySfTPLx7v6N+UbDSqrq8Ul2zHDTnN9PckF3v2K+UbFZVZ2bZJ/uPmTy/KXRm270qupdSQ5OspeP+I9DVd2W5LYM30/fnYw9JckVSQ7u7g/NMz7gZ+T44yfHHz85/rjJ8ReTHH981nOO7yQ4jEhV7Z7k0iTXJ3nhnMNhC7r7s5MvP1ZVG5O8s6rO6O4vzzMukqp6bIbTVU+pql0nwztO/tylqu7o7h+u/Grm7KIkz0uyb9xBfiw2Jbl2c3I88bEMSfP+SZY2QQZYS3L8xSDHHy85/kKT44/Pus3xFcFhJKpqxyTvz3ADlud09w/mHBKrszlZfmQSCfL8PSbJ/ZP89QrXvpbk7Ul+c7tGxNbyEbXx+GJ+9rHwaZXESR6AVZDjLyw5/rjI8RefHH881m2OrwgOI1BVG5K8O8Mv9yd197fnHBKr96TJn1+ZaxRs9rEkT5sZe1aS303y7Dh9MGZHJNmY5KvzDoSfen+SP6iqPbt742TsKRn+Efr5+YUFsBjk+AtNjj8ucvzFJccfn3Wb4yuCrwOT0wfPnjx9eJKdq+q5k+f/s7tvmU9kTHlzhj06McnuVXXg1LXPdfet8wmLaVX1gSSXJ/nbJHdkSI7/Q5J3+ZjkOEx+iX9kemzSgzNJPqpf4DhU1Xsy3DDnCxlumnPU5HGCXoGjcl6SE5JcUlWvzXDTnNcnuby7PzbXyAA5/mKQ4y8AOf74yfEXgxx/YazbHF8RfH14aIYTCNM2P39kkuu2azSs5BmTP89e4Zo9Go9PZ7iL8r5JfpzhxMHJSc6dX0iwkK7J0Ndx7wwfu7sqyYu7+/y5RsWddPdNVfX0JP8lyYUZ+gReWsNL9QAABPdJREFUnOSVcw0M2EyOP35y/MUgx4e1IcdfAOs5x69ubXkAAAAAAFhO95t3AAAAAAAAsK0oggMAAAAAsLQUwQEAAAAAWFqK4AAAAAAALC1FcAAAAAAAlpYiOAAAAAAAS0sRHGBkqurwqvpwVd1YVbdW1d9V1R9W1Z6T611Vx69yrXdO5r98C9f3qarzq+r6qvpRVf1DVV1cVU+ZmffSqvpMVX2/qjZV1eeq6syZOb2Fx5e2di0AAFgmcnyA+VIEBxiRqjojybuTXJvkRUmekeSsJIcm+bOtXOtBSf715OnRK1zfLcknkjw2yclJDkny+0l+kuQJU/NOTvK2JJclOTzJi5NcnOTXV3jbMyavnX48916uBQAAC0+ODzB/1d3zjgGAJFV1aJL3JXl5d//XmWs7JHlGd19aVZ3kt7v7nHtY74gkFyX5UJKnJXl4d98wdf3fJXlrkod197dnXls9+QVRVV9P8j+6+7gtzZk8v8e4VrsWAAAsAzm+HB8YByfBAcbjlUk+O5scJ0l339Hdl27lekcn+XqS4zP8vH/ezPVdk9yW5P+t8H49M++Ge5izWmu5FgAAjJ0cH2AEFMEBRqCq7p/kiUk+sEbr7ZTkOUn+sruvTvLZ3PXjkp9N8sAk51fVP6uqLf1O+GyS366ql1TVHvfw1verqg0zj+l1t2YtAABYWHJ8gPFQBAcYhz0yJKvXr9F6v5HkQUkunDy/MMmBVfXIzRO6+0MZehEeleTKJDdW1Xuq6uCZtY5LcnOSdyT5TlX9bVWdXlU7r/C+Zye5feYxfepla9YCAIBFJscHGAlFcIBxWauPDB6d5Nru/tTk+YWTtZ9/pzfrPinJP0ryqiQfSfKsJH9VVcdOzflCkl/OcGObNyepJK9OcmVV/dzM+74xyT+feZx2L9cCAIBlIMcHmDNFcIBx+G6SW5M84r4uVFV7Jjk4ySVVtWtV7Zrk+0k+neQFs/O7+0vd/cfd/etJ9knyN0leW1U1NefW7r6ku4/v7v2T/GaSxyR5+cxy13f3lTOP62beb7VrAQDAIpPjy/GBkVAEBxiB7r49yf9O8sw1WO7IJBuSnJhk09TjV5I8rqoedzdxbEzy50l2S/LQu5n39gw329nvvga7lmsBAMBYyPHl+MB4KIIDjMefJDmgql4ye6Gq7ldVz1rlOkcn+WKSp808npWhf9/zJ2vutYXXPybDiZXvTebdJVGevHaXJN9aZUybX7dmawEAwAKQ4wOMwIZ5BwDAoLsvqaozk7y9qp6U5OIMN5jZL8mxSa7Lz+4s/0+r6rkzS3wnyZeTPDnJyd39kdn3qKoPZEigT03ykqr6N0n+Isnnk9w/yUFJXpHkLd39o8nL/k9VXZzkr5J8O8PHKX8nyS1J3jnzFvtW1YF3/av1J+/FWgAAsNDk+HJ8YByqe63uzwDAWqiqI5Icn+TxSR6cITF+X5I/7u4bqmpLP7ivSPL+JK9Psk93f22FtZ+X5F1JDszQQ/C4DCdI9k5yR4YE+21J/qy7fzx5zXFJDkvyuCS7J7khyceTnN7dV0+tvaW47ujuDVuzFgAALBM5PsB8KYIDAAAAALC09AQHAAAAAGBpKYIDAAAAALC0FMEBAAAAAFhaiuAAAAAAACwtRXAAAAAAAJaWIjgAAAAAAEtLERwAAAAAgKWlCA4AAAAAwNJSBAcAAAAAYGn9f5Q3eedHdHz8AAAAAElFTkSuQmCC\n"},"metadata":{"needs_background":"light"}},{"output_type":"stream","text":"NearMiss Version 1\nClass=1, n=2053 (44.679%)\nClass=2, n=783 (17.040%)\nClass=4, n=165 (3.591%)\nClass=6, n=378 (8.226%)\nClass=5, n=1126 (24.505%)\nClass=3, n=90 (1.959%)\n\n\nNearMiss Version 2\nClass=6, n=670 (14.581%)\nClass=2, n=1453 (31.621%)\nClass=4, n=212 (4.614%)\nClass=1, n=1894 (41.219%)\nClass=5, n=242 (5.267%)\nClass=3, n=124 (2.699%)\n","name":"stdout"}]},{"metadata":{},"cell_type":"markdown","source":"## Both have given non-skewed values for this dataset, hence we can conclude that for multidimensional datasets using Undersampling techniques are better."},{"metadata":{"trusted":true},"cell_type":"code","source":"","execution_count":null,"outputs":[]}],"metadata":{"kernelspec":{"name":"python3","display_name":"Python 3","language":"python"},"language_info":{"name":"python","version":"3.7.6","mimetype":"text/x-python","codemirror_mode":{"name":"ipython","version":3},"pygments_lexer":"ipython3","nbconvert_exporter":"python","file_extension":".py"}},"nbformat":4,"nbformat_minor":4}