ml_kaggle-home-loan-credit-risk-feateng.py (Source) 

  1
  2
  3
  4
  5
  6
  7
  8
  9
 10
 11
 12
 13
 14
 15
 16
 17
 18
 19
 20
 21
 22
 23
 24
 25
 26
 27
 28
 29
 30
 31
 32
 33
 34
 35
 36
 37
 38
 39
 40
 41
 42
 43
 44
 45
 46
 47
 48
 49
 50
 51
 52
 53
 54
 55
 56
 57
 58
 59
 60
 61
 62
 63
 64
 65
 66
 67
 68
 69
 70
 71
 72
 73
 74
 75
 76
 77
 78
 79
 80
 81
 82
 83
 84
 85
 86
 87
 88
 89
 90
 91
 92
 93
 94
 95
 96
 97
 98
 99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
 
#!/usr/bin/env python3
# -*- coding: utf-8 -*-
"""
Created on Tue Jan 29 22:15:16 2019

@author: randlow
"""

import sys
import pandas as pd
import pickleshare as pkl
import os


from pathlib import Path

home = str(Path.home())
if sys.platform == 'linux':
    inputDir = "/datasets/kaggle/home-credit-default-risk" # linux
elif sys.platform == 'win32':
    inputDir = "\datasets\kaggle\home-credit-default-risk" # windows

fullDir = home+inputDir
os.chdir(fullDir)

storeDir = fullDir+'/pickleshare'

db = pkl.PickleShareDB(storeDir)

# In[50]"
print(db.keys())
x1 = db['df_app_test_align']

# Based on our empirical analysis of just analyzing correlations between our target variable and the feature variables, we can perform feature engineering.  Typically feature engineering means we perform operations such as:
# 
# 1. Polynomial features: THis includes all interactions and powers of each feature variable
# 2. Expert knowledge features:
# In[46]:

df_app_train = pd.read_csv('application_train.csv',index_col=0)
df_app_test = pd.read_csv('application_test.csv',index_col=0)


# In[47]:


print(var_pos_corr[0:10])

imp_var = var_pos_corr[0:4]
print(imp_var)


# In[48]:


poly_features_train = df_app_train[imp_var]
poly_features_test = df_app_test[imp_var]

poly_target_train = df_app_train['TARGET']

poly_features_train.columns


# ## Imputing NaN points

# In[49]:


imputer = preprocessing.Imputer(strategy='median')
poly_features_train = imputer.fit_transform(poly_features_train) # fitting means we find the median then apply
poly_features_test = imputer.transform(poly_features_test) # we only transform, as we want to use the fit used on the training dataset


# ## Creating polynomial features

# In[50]:


poly_transformer = preprocessing.PolynomialFeatures(degree=3)

poly_transformer.fit(poly_features_train)

poly_features_train = poly_transformer.transform(poly_features_train)
poly_features_test = poly_transformer.transform(poly_features_test)

print('Polynomial features: {}'.format(poly_features_train.shape))


# In[51]:


poly_transformer.get_feature_names()[:15]


# In[52]:


poly_transformer.get_feature_names(input_features=imp_var)


# Since we now have a bigger dataset with additional artificially created features, let's evaluate if these new features have higher correlation than the original set of features. `poly_features_train` is an array, so we need to create a DataFrame out of it, and add the `Target` column to it.

# In[53]:


df_poly_features_train = pd.DataFrame(poly_features_train, columns = poly_transformer.get_feature_names(input_features=imp_var))
df_poly_features_train['TARGET'] = poly_target_train


# In[54]:


poly_corrs = df_poly_features_train.corr()['TARGET'].sort_values()
print('+ve correlations:\n{}'.format(poly_corrs.tail(20)))
print('-ve correlations:\n{}'.format(poly_corrs.head(20)))


# Now that we have a new `df_poly_features_train` that includes all the existing features, and polynomial features, we need to add in the `SK_ID_CURR` column too

# In[59]:


df_app_train.index.values


# In[56]:


df_poly_features_train['SK_ID_CURR'] = df_app_train['SK_ID_CURR'] # Now that we have
#df_app_poly_train = df_app_train.merge(df_poly_features_train, on = 'SK_ID_CURR', how='left')


# In[ ]:





# ## Add new features to the `test` dataset
# 

# In[ ]:


df_poly_features_test = pf.DataFrame(poly_features_test, columns = poly_transformer.get_feature_names(input_features=imp_var))
df_poly_features_test['SK_ID_CURR'] = df_app_test['SK_ID_CURR']
df_app_poly_test = df_app_test.merge(df_poly_features_test, on='SK_ID_CURR', how='left')


# ## Align the `train` and `test` datasets

# In[ ]:


df_app_poly_train, df_app_poly_test = app_train_poly_train.align(df_app_poly_test, join='inner',axis=1) 


# haha

# hehe

# In[61]:


get_ipython().system('jupyter nbconvert --to script ml_kaggle_home-loan-credit-risk.ipynb')


# In[ ]: