As for now, we can create only shap values for models with ecy as a feature. Base margine is not taken into account in shap package
In [1]:
import pandas as pd
import numpy as np
import pickle
import xgboost as xgb
import os
import shap
In [2]:
ModelsDir = '/home/kate/Research/Property/Models/'
DataDir = '/home/kate/Research/Property/Data/'
In [3]:
Model = 'wc_class_f_ecy_XGB'
In [4]:
dataset = pd.read_csv(DataDir+'property_water_claims_non_cat_fs.csv', error_bad_lines=False, index_col=False)
In [5]:
featureset_shap = [
'roofcd_encd_shap_value',
'sqft_shap_value',
'usagetype_encd_shap_value',
'yearbuilt_shap_value',
'water_risk_3_blk_shap_value',
'landlordind_shap_value',
'multipolicyind_shap_value',
'cova_deductible_shap_value',
'cova_limit_shap_value',
'ecy_shap_value'
]
In [6]:
featureset = [
'roofcd_encd',
'sqft',
'usagetype_encd',
'yearbuilt',
'water_risk_3_blk',
'landlordind',
'multipolicyind',
'cova_deductible',
'cova_limit',
'ecy'
]
In [7]:
kfold = 5
In [8]:
#
X=dataset[featureset]
DX = xgb.DMatrix(X.values)
Creating shap values for testing, prediction and training datasets¶
In [ ]:
for i in range(0,kfold):
dataset_shap_values = pd.DataFrame()
ModelName=Model+"_%s"%i
xgb_model_file='%s%s.model'%(ModelsDir,ModelName)
print('Processing model %s, fold %s...'%(Model,i))
xgb_model = pickle.load(open(xgb_model_file, 'rb'))
explainer = shap.TreeExplainer(xgb_model)
#dataset explaining
shap_values = explainer.shap_values(DX)
df_shap_values = pd.DataFrame(data=shap_values, columns=featureset_shap)
df_shap_values['original_output_value'] = df_shap_values.sum(axis=1)
df_shap_values['expected_value'] = explainer.expected_value
df_shap_values['output_value'] = df_shap_values['expected_value'] + df_shap_values['original_output_value']
df_shap_values['modeldata_id'] = dataset['modeldata_id']
df_shap_values['cal_year'] = dataset['cal_year']
df_shap_values['ModelName'] = Model
df_shap_values['fold'] = i
for f in featureset:
df_shap_values[f] = dataset[f]
df_shap_values = df_shap_values[['ModelName','fold','modeldata_id','cal_year']+featureset_shap+['original_output_value','expected_value','output_value']+featureset]
dataset_shap_values = dataset_shap_values.append(df_shap_values)
#Saving
dataset_shap_values.to_csv('%sfrequency_shap_values_%s.csv'%(DataDir,ModelName),header=True,index=False)
In [12]:
#shap.initjs()
In [13]:
#shap.summary_plot(shap_values, X)
In [18]:
Model = 'wc_Poisson_f_ecy_XGB'
In [19]:
for i in range(0,kfold):
dataset_shap_values = pd.DataFrame()
ModelName=Model+"_%s"%i
xgb_model_file='%s%s.model'%(ModelsDir,ModelName)
print('Processing model %s, fold %s...'%(Model,i))
xgb_model = pickle.load(open(xgb_model_file, 'rb'))
explainer = shap.TreeExplainer(xgb_model)
#dataset explaining
shap_values = explainer.shap_values(DX)
df_shap_values = pd.DataFrame(data=shap_values, columns=featureset_shap)
df_shap_values['original_output_value'] = df_shap_values.sum(axis=1)
df_shap_values['expected_value'] = explainer.expected_value
df_shap_values['output_value'] = df_shap_values['expected_value'] + df_shap_values['original_output_value']
df_shap_values['modeldata_id'] = dataset['modeldata_id']
df_shap_values['cal_year'] = dataset['cal_year']
df_shap_values['ModelName'] = Model
df_shap_values['fold'] = i
for f in featureset:
df_shap_values[f] = dataset[f]
df_shap_values = df_shap_values[['ModelName','fold','modeldata_id','cal_year']+featureset_shap+['original_output_value','expected_value','output_value']+featureset]
dataset_shap_values = dataset_shap_values.append(df_shap_values)
#Saving
dataset_shap_values.to_csv('%sfrequency_shap_values_%s.csv'%(DataDir,ModelName),header=True,index=False)