-
Notifications
You must be signed in to change notification settings - Fork 3
/
feature_importance.py
299 lines (241 loc) · 12.3 KB
/
feature_importance.py
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
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
# -*- coding: utf-8 -*-
import random
import numpy as np
from keras.models import Sequential, Model
from keras.layers import Dense, Dropout, Lambda, Input, Subtract, Add
from keras.optimizers import Adam
from keras.utils import to_categorical
from keras import losses
import keras.backend as K
import pandas as pd
import seaborn as sns
import matplotlib.pyplot as plt
from sklearn.model_selection import train_test_split
import shap
from scipy import stats
from lib.treatments import hypertension_treatments, dm_treatment, ascvd_treatment, hypertension_treatments_history, dm_treatment_history, framingham_ascvd_risk
from lib.DQN import DQNAgent
from lib.next_state_cols import next_state_cols
target_treatment = None
action_size = 100
excluded = ['study_id', 'encounter_dt_ran']
data_path = '~/Research/PHD/project/Hua Zheng/previous code/cleaned_EHR_treatment_param_lab_test_final_3diseases-v6.csv'
# '~/Research/PHD/project/Hua Zheng/previous code/cleaned_EHR_treatment_param_lab_test_final_3diseases_cvd_encounter_diagnosis.csv'
data = pd.read_csv(data_path)
data = data.drop(['egfr_mdrd_african_american_min', 'egfr_mdrd_african_american_max', 'egfr_mdrd_african_american',
'egfr_mdrd_non_african_american', 'egfr_mdrd_non_african_american_max',
'egfr_mdrd_non_african_american_min', 'next_egfr_mdrd_african_american',
'next_egfr_mdrd_african_american_max', 'next_egfr_mdrd_african_american_min',
'next_egfr_mdrd_non_african_american', 'next_egfr_mdrd_non_african_american_max',
'next_egfr_mdrd_non_african_american_min'], axis=1)
data = data.dropna()
if diagnosis_reward:
label_cols = list(data.columns[116:159])
else:
label_cols = list(data.columns[117:160])
full_label_cols = label_cols
if target_treatment != None:
label_cols = treatment_ctg[target_treatment]
action_size = len(label_cols)
#data = data.loc[data[label_cols].sum(axis=1, skipna=True) != 0,]
data = data.loc[(data[label_cols].sum(axis=1, skipna=True) != 0) & (
data[set(full_label_cols) - set(label_cols)].sum(axis=1, skipna=True) == 0),]
target = data[label_cols].apply(lambda x: hash_to_action(x), axis=1)
data['target'] = target
counter = collections.Counter(target)
target_set = list([i[0] for i in counter.most_common(action_size)])
target_replacement = dict(zip(iter(target_set), range(action_size)))
target_column_renames = ['target' + str(i) for i in range(action_size)]
data = data[data.target.apply(lambda x: x in target_set)]
data['target'] = data['target'].replace(target_replacement)
reward_cols = ['reward']
# %%
state_cols = [s[5:] for s in next_state_cols]
# state_cols = list(set(data.columns) - set(full_label_cols) - {'target'} - set(
# ['reward', 'reward_bp', 'reward_ascvd', 'reward_diabetes', 'risk_ascvd', 'next_risk_ascvd']) - set(
# ['study_id', 'encounter_dt_ran']) - set(next_state_cols))
if diagnosis_reward:
state_cols = state_cols - (['CVD', 'days_to_CVD'])
# next_state_cols = ['next_' + s for s in list(state_cols)]
patients_column = data[['study_id', 'encounter_dt_ran']]
#data = data.drop(excluded, axis=1)
_temp = data.drop('target', axis=1).max(skipna=True) - data.drop('target', axis=1).min(skipna=True)
_temp[data.drop('target', axis=1).columns[(_temp == 0).values]] = 1.0
normalized_df = (data.drop(['target'], axis=1) - data.drop('target', axis=1).min(skipna=True)) / _temp
# constant_variable_df = data.drop('target', axis=1).max(skipna=True) - data.drop('target', axis=1).min(skipna=True)
# dropped_state = normalized_df.columns[(constant_variable_df == 0).values]
# normalized_df[dropped_state] = data[dropped_state]
normalized_df['reward'] = (data['reward'] - data['reward'].min(skipna=True)) / (data['reward'].max() / 3)
# next_state_cols = set(next_state_cols) - set(dropped_state)
# state_cols = set(state_cols) - set(dropped_state)
ohe = to_categorical(data['target'], action_size)
for i, col in enumerate(target_column_renames):
normalized_df[col] = ohe[:, i]
normalized_df[['study_id', 'encounter_dt_ran']] = patients_column
train, test = train_test_split(normalized_df, test_size=0.4, random_state=2019)
train2, test2 = train_test_split(data, test_size=0.4, random_state=2019)
# x_train = train[state_cols]
# y_train_class = train[label_cols]
# y_train_category = train[['RX_CATEGORY']]
# x_test = test[state_cols]
# y_test_class = test[label_cols]
state_size = len(state_cols)
action_size = len(target_column_renames)
agent = DQNAgent(state_size, action_size, target_column_renames, state_cols, reward_cols, next_state_cols)
agent.model = agent.save("./3d256-512-256-episodes20000--dqn-mse-target_treatmentNone")
from IPython.display import display
# agent.model = load_model('model-5rounds-l3-tanh-lr0004.h5')
x_train = train[state_cols]
x_test = test[state_cols]
''' obtain AI feature importance
'''
background = x_train.values[np.random.choice(x_train.shape[0], 100, replace=False)]
e = shap.DeepExplainer(agent.model, background)
shap_values = e.shap_values(x_test.values[:100])
# takes long time thus save first
np.save('shap_values',shap_values)
''' obtain clinicians' feature importance
'''
model = xgb.XGBClassifier(max_depth=2,
n_estimators=300,
num_class= 100,
colsample_bytree=0.1,
subsample=0.8,
learning_rate=0.1,
objective='multi:softmax',
gpu_id=0,
eval_metric=['mlogloss'],
n_jobs=4)
model.fit(np.array(train[state_cols]), np.array(train2['target']),
eval_set=[(np.array(test[state_cols]), np.array(test2['target']))],
verbose=1,
early_stopping_rounds=4)
mybooster = model.get_booster()
model_bytearray = mybooster.save_raw()[4:]
def myfun(self=None):
return model_bytearray
mybooster.save_raw = myfun
# Shap explainer initilization
shap_ex = shap.TreeExplainer(mybooster)
e = shap_ex.shap_values(np.array(test[state_cols])[np.random.choice(test.shape[0], 5000, replace=False)])
clinician_shap_values = pd.DataFrame(np.abs(np.array(e)).mean(0))
clinician_shap_values = np.abs(clinician_shap_values)
k_clinician=pd.DataFrame(clinician_shap_values.mean()).reset_index()
AI_shap_values = np.load('/Users/hua.zheng/PycharmProjects/ReinforcementLearning/results/shap_values.npy')
AI_shap_values = pd.DataFrame(np.abs(np.array(AI_shap_values)).mean(0))
AI_shap_values = np.abs(AI_shap_values)
k=pd.DataFrame(AI_shap_values.mean()).reset_index()
data = pd.DataFrame({'feature':state_cols,'AI': k[0],'clincian': k_clinician[0]})
data = pd.read_csv('raw_feature_importance.csv',index=None)
data.drop('Unnamed: 0',inplace=True,axis=1)
biomarker_impact = data.loc[(~data['feature'].str.contains('hist')) & (~data['feature'].str.contains('cur')),]
treatment_hist = data.loc[(data['feature'].str.contains('hist')) | (data['feature'].str.contains('cur')),]
cleaned_biomarker_impact = biomarker_impact.loc[(~biomarker_impact['feature'].str.contains('max')) & (~biomarker_impact['feature'].str.contains('min')),]
feature_list_ = []
biomarker_list_AI = []
biomarker_list_clincian = []
for bio in cleaned_biomarker_impact['feature'].values:
if bio == 'hdl_cholesterol':
feature_list_.append('HDL')
elif bio == 'ldl_cholesterol':
feature_list_.append('LDL')
elif bio == 'cholesterol,_total':
feature_list_.append('TC')
elif bio == 'bp_systolic':
feature_list_.append('Systolic BP')
elif bio == 'bp_diastolic':
feature_list_.append('Diastolic BP')
elif bio == 'time_last_vist':
feature_list_.append('Time since last vist')
elif bio == 'bmi':
feature_list_.append('BMI')
elif bio == 'hemoglobin_a1c':
feature_list_.append('HbA1c')
elif bio == 'sex_male':
feature_list_.append('Sex(male)')
elif bio == 'creatinine':
feature_list_.append('Creatinine')
elif bio == 'triglycerides':
feature_list_.append('Triglycerides')
elif bio == 'age':
feature_list_.append('Age')
elif bio == 'smoke':
feature_list_.append('Smoke')
elif bio == 'time_to_first_visit':
feature_list_.append('Time since first visit')
else:
feature_list_.append(bio.replace('_', ' '))
biomarker_list_AI.append(biomarker_impact.loc[biomarker_impact['feature'].str.contains(bio),].mean()['AI'])
biomarker_list_clincian.append(biomarker_impact.loc[biomarker_impact['feature'].str.contains(bio),].mean()['clincian'])
feature_list_.append('race')
biomarker_list_AI.append(biomarker_impact.loc[biomarker_impact['feature'].str.contains('race'),].mean()['AI'])
biomarker_list_clincian.append(biomarker_impact.loc[biomarker_impact['feature'].str.contains('race'),].mean()['clincian'])
feature_list_.append('Recent Therapies')
biomarker_list_AI.append(treatment_hist['AI'].mean() / 1.5)
biomarker_list_clincian.append(treatment_hist['clincian'].mean())
feature_importance = pd.DataFrame({'variables': feature_list_, 'AI': biomarker_list_AI, 'clinician': biomarker_list_clincian})
feature_importance.iloc[16,1] = feature_importance.iloc[16,]['AI'] / 1.5
feature_importance = feature_importance.loc[(~feature_importance['variables'].str.contains('race')) & (~feature_importance['variables'].str.contains('race')),]
# feature_importance.to_csv('feature_importance3.csv')
# data.to_csv('raw_feature_importance.csv')
#
# import seaborn as sns
# sns.set(style="whitegrid")
feature_group = ['biomarker', 'biomarker', 'biomarker','biomarker','biomarker',
'treatment history', 'biomarker', 'biomarker', 'demographics',
'demographics','biomarker','demographics','biomarker','treatment history','treatment history']
feature_importance['feature_group'] =feature_group
feature_importance.sort_values(by='feature_group',inplace=True)
# Make the PairGrid
feature_importance_temp = feature_importance.copy()
feature_importance_temp['AI'] = feature_importance_temp['AI'].values * 100
g = sns.PairGrid(feature_importance_temp,
x_vars=feature_importance.columns[1:3], y_vars=["variables"],
height=5, aspect=1)
# Draw a dot plot using the stripplot function
g.map(sns.stripplot, size=10, orient="h",
# palette="ch:s=1,r=-.1,h=1_r",
linewidth=1,
edgecolor="w")
# Use the same x axis limits on all columns and add better labels
# g.set(xlim={(0, 0.02), xlabel={"Crashes", '1'}, ylabel="")
# for ax, title in zip(g.axes.flat, titles):
#
# # Set a different title for each axes
# ax.set(title=title)
#
# # Make the grid horizontal instead of vertical
# ax.xaxis.grid(False)
# ax.yaxis.grid(True)
sns.despine(left=True, bottom=True)
plt.savefig('feature_importance.svg')
plt.show()
temp_feature_imp = feature_importance.groupby('feature_group').apply(lambda x: x.sort_values(by='AI',ascending=True))
temp_feature_imp.reset_index()
temp_feature_imp['AI'] = temp_feature_imp['AI'] * temp_feature_imp['clinician'].sum() / temp_feature_imp['AI'].sum()
temp_feature_imp.plot(kind= 'barh' , rot=0, x='variables', figsize=(20,10), fontsize=20)
cmap = plt.get_cmap("tab20c")
outer_colors = cmap(6)
in_colors = cmap(0)
feature_importance.sort_values('AI',ascending=True)[['variables','AI']].plot(kind= 'barh' , rot=0, x='variables', figsize=(12,12), fontsize=20,color=in_colors, alpha=0.9,zorder=2)
plt.show()
# width = 0.4
# temp_feature_imp.AI.plot(kind='barh', color='red', ax=ax, width=width, position=0,x='variables')
# temp_feature_imp.clinician.plot(kind='barh', color='blue', ax=ax2, width=width, position=1,x='variables')
#
# ax2 = ax.twinx()
# seaxs = ax.secondary_xaxis('top', functions=rho2m)
plt.savefig('feature_imp1-ai.png')
plt.savefig('feature_imp2-ai.pdf')
plt.savefig('feature_imp3-ai.svg')
feature_importance.sort_values('clinician',ascending=True)[['variables','clinician']].plot(kind= 'barh' , rot=0, x='variables', figsize=(12,12), fontsize=20,color=outer_colors, alpha=0.9, zorder=2)
plt.show()
# width = 0.4
# temp_feature_imp.AI.plot(kind='barh', color='red', ax=ax, width=width, position=0,x='variables')
# temp_feature_imp.clinician.plot(kind='barh', color='blue', ax=ax2, width=width, position=1,x='variables')
#
# ax2 = ax.twinx()
# seaxs = ax.secondary_xaxis('top', functions=rho2m)
plt.savefig('feature_imp1-ai.png')
plt.savefig('feature_imp2-ai.pdf')
plt.savefig('feature_imp3-ai.svg')