RUN PREDICTION (Gradient Boosting Model)



In [16]:

    
import pandas as pd
import pickle

LOAD TEST DATA



In [10]:

    
df = pd.read_csv('https://s3.amazonaws.com/marweezys-bucket/all_state_insurance_prediction/test.csv')
ids = list(df['id'])
df.drop(labels='id', axis=1, inplace=True)
one_hot_df = pd.get_dummies(df.ix[:,:])
one_hot_df['cat116_W'] = 0
one_hot_df.head()









    Out[10]:






  
    
      
      cont1
      cont2
      cont3
      cont4
      cont5
      cont6
      cont7
      cont8
      cont9
      cont10
      ...
      cat116_MX
      cat116_N
      cat116_O
      cat116_Q
      cat116_R
      cat116_S
      cat116_T
      cat116_U
      cat116_Y
      cat116_W
    
  
  
    
      0
      0.321594
      0.299102
      0.246911
      0.402922
      0.281143
      0.466591
      0.317681
      0.61229
      0.34365
      0.38016
      ...
      0.0
      0.0
      0.0
      0.0
      0.0
      0.0
      0.0
      0.0
      0.0
      0
    
    
      1
      0.634734
      0.620805
      0.654310
      0.946616
      0.836443
      0.482425
      0.443760
      0.71330
      0.51890
      0.60401
      ...
      0.0
      0.0
      0.0
      0.0
      0.0
      0.0
      0.0
      0.0
      0.0
      0
    
    
      2
      0.290813
      0.737068
      0.711159
      0.412789
      0.718531
      0.212308
      0.325779
      0.29758
      0.34365
      0.30529
      ...
      0.0
      0.0
      0.0
      0.0
      0.0
      0.0
      0.0
      0.0
      0.0
      0
    
    
      3
      0.268622
      0.681761
      0.592681
      0.354893
      0.397069
      0.369930
      0.342355
      0.40028
      0.33237
      0.31480
      ...
      0.0
      0.0
      0.0
      0.0
      0.0
      0.0
      0.0
      0.0
      0.0
      0
    
    
      4
      0.553846
      0.299102
      0.263570
      0.696873
      0.302678
      0.398862
      0.391833
      0.23688
      0.43731
      0.50556
      ...
      0.0
      0.0
      0.0
      0.0
      0.0
      0.0
      0.0
      0.0
      0.0
      0
    
  

5 rows × 1117 columns

LOAD MODEL



In [11]:

    
with open('GB_feats.plk','rb') as f:
    feats = pickle.load(f)
with open('GB_model.plk','rb') as f:
    model = pickle.load(f)

MAKE PREDICTIONS



In [13]:

    
X = one_hot_df[feats]
y_pred = model.predict(X)

SAVE/SHOW PREDICTIONS



In [19]:

    
GB_d = {}
GB_d['id'] = ids
GB_d['loss'] = y_pred
GB_df = pd.DataFrame(GB_d)
GB_df.to_csv('GB_preds.csv', sep=',', index=False)
GB_df.head()









    Out[19]:






  
    
      
      id
      loss
    
  
  
    
      0
      4
      1579.628558
    
    
      1
      6
      2265.133652
    
    
      2
      9
      9989.328296
    
    
      3
      12
      6191.534165
    
    
      4
      15
      1151.990087

	cont1	cont2	cont3	cont4	cont5	cont6	cont7	cont8	cont9	cont10	...
0	0.321594	0.299102	0.246911	0.402922	0.281143	0.466591	0.317681	0.61229	0.34365	0.38016	...
1	0.634734	0.620805	0.654310	0.946616	0.836443	0.482425	0.443760	0.71330	0.51890	0.60401	...
2	0.290813	0.737068	0.711159	0.412789	0.718531	0.212308	0.325779	0.29758	0.34365	0.30529	...
3	0.268622	0.681761	0.592681	0.354893	0.397069	0.369930	0.342355	0.40028	0.33237	0.31480	...
4	0.553846	0.299102	0.263570	0.696873	0.302678	0.398862	0.391833	0.23688	0.43731	0.50556	...

	id	loss
0	4	1579.628558
1	6	2265.133652
2	9	9989.328296
3	12	6191.534165
4	15	1151.990087