-
Notifications
You must be signed in to change notification settings - Fork 0
/
Transformer.py
365 lines (302 loc) · 20.5 KB
/
Transformer.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
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
import pandas as pd
from datetime import date, datetime
from sklearn.base import BaseEstimator, TransformerMixin
from sklearn.pipeline import Pipeline
######################################################################
###------Load and Merge------#########################################
######################################################################
# load the data from csv to pandas dataframe
provider = pd.read_csv("data/Train-1542865627584.csv")
beneficiary = pd.read_csv("data/Train_Beneficiarydata-1542865627584.csv")
inpatient = pd.read_csv("data/Train_Inpatientdata-1542865627584.csv")
outpatient = pd.read_csv("data/Train_Outpatientdata-1542865627584.csv")
# union/concat the inpatient and outpatient data
concat_df=pd.concat([inpatient, outpatient],axis=0)
merge_bene_df=concat_df.merge(beneficiary, on='BeneID', how='left')
merge_provider_df=merge_bene_df.merge(provider, on = 'Provider', how ='left')
diagnosis_code_columns = ['ClmDiagnosisCode_1', 'ClmDiagnosisCode_2', 'ClmDiagnosisCode_3', 'ClmDiagnosisCode_4', 'ClmDiagnosisCode_5',
'ClmDiagnosisCode_6', 'ClmDiagnosisCode_7', 'ClmDiagnosisCode_8', 'ClmDiagnosisCode_9', 'ClmDiagnosisCode_10' ]
procedure_code_columns = ['ClmProcedureCode_1', 'ClmProcedureCode_2', 'ClmProcedureCode_3',
'ClmProcedureCode_4', 'ClmProcedureCode_5', 'ClmProcedureCode_6']
train_top_15_diag_codes = ['4019', '25000', '2724', 'V5869', '4011', '42731', 'V5861', '2720', '2449',
'4280', '53081', '41401', '496', '2859', '41400', 'Other']
train_top_15_proc_codes = ['4019.0', '9904.0', '2724.0', '8154.0', '66.0', '3893.0', '3995.0', '4516.0',
'3722.0', '8151.0', '8872.0', '9671.0','4513.0','5849.0', '9390.0', 'Other']
fraction_column_list= ['ChronicCond_Alzheimer','ChronicCond_Heartfailure','ChronicCond_KidneyDisease','ChronicCond_Cancer',
'ChronicCond_ObstrPulmonary','ChronicCond_Depression','ChronicCond_Diabetes','ChronicCond_IschemicHeart','ChronicCond_Osteoporasis',
'ChronicCond_rheumatoidarthritis','ChronicCond_stroke', 'RenalDiseaseIndicator', 'Deceased', 'Gender', 'Race']
#######################################################################################
####################------Transformer Classes------####################################
#######################################################################################
class DateTransform(BaseEstimator, TransformerMixin):
# initializer
def __init__(self, start, end, newColumn):
# save the features list internally in the class
self.start = start
self.end = end
self.newColumn = newColumn
def fit(self, X, y=None):
return self
def transform(self, X, y=None):
X[self.newColumn] = self.convertDateToPeriod(X, self.start, self.end)
X[self.newColumn] = X[self.newColumn].fillna(0)
return X
def fit_transform(self, X, y=None):
self.fit(X)
return self.transform(X)
def convertDateToPeriod(self, df, startDate, endDate):
return (pd.to_datetime(df[endDate]) - pd.to_datetime(df[startDate])).dt.days + 1
class AgeTransform(BaseEstimator, TransformerMixin):
# initializer
def __init__(self, dob, dod, ageColumn, deceasedColumn):
# save the features list internally in the class
self.dob = dob
self.dod = dod
self.ageColumn = ageColumn
self.deceasedColumn = deceasedColumn
def fit(self, X, y=None):
return self
def transform(self, X, y=None):
X[self.ageColumn] = X.apply(lambda x: self.calculateAge(dob = x[self.dob], dod = x[self.dod], calulationDate = '2009-12-01'), axis = 1)
X[self.deceasedColumn] = X[self.dod].apply(lambda x : 0 if pd.isna(x) else 1)
return X
def fit_transform(self, X, y=None):
self.fit(X)
return self.transform(X)
def calculateAge(self, dob, dod, calulationDate):
born = datetime.strptime(dob, "%Y-%m-%d").date()
if not pd.isna(dod):
calulationDate = datetime.strptime(dod, "%Y-%m-%d").date()
else:
calulationDate = datetime.strptime(calulationDate, "%Y-%m-%d").date()
return calulationDate.year - born.year - ((calulationDate.month, calulationDate.day) < (born.month, born.day))
class CodeCountTransform(BaseEstimator, TransformerMixin):
# initializer
def __init__(self, colunmsToCount, newColumn):
# save the features list internally in the class
self.colunmsToCount = colunmsToCount
self.newColumn = newColumn
def fit(self, X, y=None):
return self
def transform(self, X, y=None):
X[self.newColumn] = self.countCodeNumber(X, self.colunmsToCount)
return X
def fit_transform(self, X, y=None):
self.fit(X)
return self.transform(X)
def countCodeNumber(self, df, colunmsToCount):
df_codes = df.loc[:, colunmsToCount]
codecount = df_codes.notnull().sum(axis=1)
return codecount
class CodeFrequencyGroupTransform(BaseEstimator, TransformerMixin):
'''code_columns: the column list containing the claim codes
new_column: list of new_columns
'''
def __init__(self, code_columns, new_columns_prefix, high, medium_high, medium, low):
# save the features list internally in the class
self.code_columns = code_columns
self.new_columns_prefix = new_columns_prefix
self.high = high
self.medium_high = medium_high
self.medium = medium
self.low = low
def fit(self, X, y=None):
self.frequency_counts = self.getTotalCodeCounts(X, self.code_columns)
self.frequency_groups = self.getFrequencyGroups(self.frequency_counts, self.high, self.medium_high,
self.medium, self.low)
def transform(self, X, y=None):
X[self.new_columns_prefix+'HighFreqCount'] = self.codeForfrequencyGroupCounts(X, self.code_columns, self.frequency_groups[0])
X[self.new_columns_prefix+'MediumHighFreqCount'] = self.codeForfrequencyGroupCounts(X, self.code_columns, self.frequency_groups[1])
X[self.new_columns_prefix+'MediumFreqCount'] = self.codeForfrequencyGroupCounts(X, self.code_columns, self.frequency_groups[2])
X[self.new_columns_prefix+'LowFreqCount'] = self.codeForfrequencyGroupCounts(X, self.code_columns, self.frequency_groups[3])
X[self.new_columns_prefix+'RareFreqCount'] = self.codeForfrequencyGroupCounts(X, self.code_columns, self.frequency_groups[4])
return X
def fit_transform(self, X, y=None):
self.fit(X)
return self.transform(X)
def mergeDictionaryWithUpdate(self, dict_1, dict_2):
for key in dict_2:
value = dict_2[key]
if key not in dict_1:
dict_1[key] = value
else:
old_value = int(dict_1[key])
dict_1[key] = old_value + value
return dict_1
'''Get total counts of each code'''
def getTotalCodeCounts(self, df, columns):
code_counts = {}
for column in columns:
value_counts = df[column].value_counts().to_dict()
code_counts = self.mergeDictionaryWithUpdate(code_counts, value_counts)
sorted_counts = dict(sorted(code_counts.items(), key=lambda item: item[1], reverse=True))
return sorted_counts
def getFrequencyGroups(self, dictionary, high, medium_high, medium, low):
high_frequency = []
medium_high_frequency = []
medium_frequency = []
low_fequency = []
rare_frequency = []
for key in dictionary:
value = dictionary[key]
if value >= high:
high_frequency.append(key)
elif value < high and value >= medium_high:
medium_high_frequency.append(key)
elif value < medium_high and value >= medium:
medium_frequency.append(key)
elif value < medium and value >= low:
low_fequency.append(key)
else:
rare_frequency.append(key)
results = []
results.extend((high_frequency, medium_high_frequency, medium_frequency, low_fequency, rare_frequency))
return results
def codeForfrequencyGroupCounts(self, df, columns, frequency_group):
df_codes = df.loc[:, columns]
codecount = df_codes.isin(frequency_group).sum(axis=1)
return codecount
class Top15OneHotTransform(BaseEstimator, TransformerMixin):
def __init__(self, column_list, top_15_codes, new_column_prefix):
self.column_list = column_list
self.top_15_codes = top_15_codes
self.new_column_prefix = new_column_prefix
self.codes_df = pd.DataFrame()
def fit(self, X, y=None):
#change codes not in top15 with 'Other'
codes_df = X[self.column_list]
for column in self.column_list:
codes_df[column] = codes_df[column].apply(lambda x : 'Other' if x not in self.top_15_codes else x )
self.codes_df = codes_df
def transform(self, X, y=None):
for code in self.top_15_codes:
column_name = self.new_column_prefix + code
X[column_name] = self.codeForNHotCounts(self.codes_df, self.column_list, code)
return X
def fit_transform(self, X, y=None):
self.fit(X)
return self.transform(X)
def codeForNHotCounts(self, df, columns, code):
df_codes = df.loc[:, columns]
codecount = (df_codes==code).sum(axis=1)
return codecount
class ProviderLevelAggregateTransform(BaseEstimator, TransformerMixin):
def __init__(self, fraction_column_list):
self.fraction_column_list = fraction_column_list
def fit(self, X, y=None):
#change codes not in top15 with 'Other'
return self
def transform(self, X, y=None):
agg_df = pd.DataFrame()
agg_df = X.groupby('Provider').agg( MedianAge = ('Age', 'median'),
MeanInscClaimAmtReimbursed = ('InscClaimAmtReimbursed', 'mean'),
MaxInscClaimAmtReimbursed = ('InscClaimAmtReimbursed', 'max'),
TotalInscClaimAmtReimbursed = ('InscClaimAmtReimbursed', 'sum'),
MeanDeductibleAmtPaid = ('DeductibleAmtPaid', 'mean'),
MaxDeductibleAmtPaid = ('DeductibleAmtPaid', 'max'),
MeanNumMonth_PartACov = ('NoOfMonths_PartACov','mean'),
MeanNumMonth_PartBCov = ('NoOfMonths_PartBCov','mean'),
MeanOPAnnualReimbursementAmt = ('OPAnnualReimbursementAmt', 'mean'),
MaxOPAnnualReimbursementAmt = ('OPAnnualReimbursementAmt', 'max'),
TotalOPAnnualReimbursementAmt = ('OPAnnualReimbursementAmt', 'sum'),
MeanOPAnnualDeductibleAmt = ('OPAnnualDeductibleAmt', 'mean'),
MaxOPAnnualDeductibleAmt = ('OPAnnualDeductibleAmt', 'max'),
TotalOPAnnualDeductibleAmt = ('OPAnnualDeductibleAmt', 'sum'),
MeanIPAnnualReimbursementAmt = ('IPAnnualReimbursementAmt', 'mean'),
MaxIPAnnualReimbursementAmt = ('IPAnnualReimbursementAmt', 'max'),
TotalIPAnnualReimbursementAmt = ('IPAnnualReimbursementAmt', 'sum'),
MeanIPAnnualDeductibleAmt = ('IPAnnualDeductibleAmt', 'mean'),
MaxIPAnnualDeductibleAmt = ('IPAnnualDeductibleAmt', 'max'),
TotalIPAnnualDeductibleAmt = ('IPAnnualDeductibleAmt', 'sum'),
MeanClaimPeriods = ('ClaimPeriod', 'mean'),
MaxHospitalDays = ('HospitalDays', 'max'),
MedianHospitalDays = ('HospitalDays', 'median'),
MeanHospitalDays = ('HospitalDays', 'mean'),
MaxDiagCodeNumPerClaim = ('DiagCodeCounts', 'max'),
MeanDiagCodeNumPerClaim = ('DiagCodeCounts', 'mean'),
MaxProcCodeNumPerClaim = ('ProcCodeCounts', 'max'),
MeanProcCodeNumPerClaim = ('ProcCodeCounts', 'mean'),
TotalDiagCodeNum = ('DiagCodeCounts', 'sum'),
TotalProcCodeNum = ('ProcCodeCounts', 'sum'),
MeanHighFreqDiagCodeNumPerClaim = ('ClmDiagHighFreqCount', 'mean'),
MeanMediumHighFreqDiagCodeNumPerClaim = ('ClmDiagMediumHighFreqCount', 'mean'),
MeanMediumFreqDiagCodeNumPerClaim = ('ClmDiagMediumFreqCount', 'mean'),
MeanLowFreqDiagCodeNumPerClaim = ('ClmDiagLowFreqCount', 'mean'),
MeanRareFreqDiagCodeNumPerClaim = ('ClmDiagRareFreqCount', 'mean'),
TotalHighFreqProcCodeNumPerClaim = ('ClmProcHighFreqCount', 'sum'),
TotalMediumHighFreqProcCodeNumPerClaim = ('ClmProcMediumHighFreqCount', 'sum'),
TotalMediumFreqProcCodeNumPerClaim = ('ClmProcMediumFreqCount', 'sum'),
TotalLowFreqProcCodeNumPerClaim = ('ClmProcLowFreqCount', 'sum'),
TotalRareFreqProcCodeNumPerClaim = ('ClmProcRareFreqCount', 'sum'),
totalDiagCode_4019 = ('DiagCode_4019', 'sum'),
totalDiagCode_25000 = ('DiagCode_25000','sum'),
totalDiagCode_2724 = ('DiagCode_2724', 'sum'),
totalDiagCode_V5869 = ('DiagCode_V5869', 'sum'),
totalDiagCode_4011 = ('DiagCode_4011', 'sum'),
totalDiagCode_42731 = ('DiagCode_42731', 'sum'),
totalDiagCode_V5861 = ('DiagCode_V5861', 'sum'),
totalDiagCode_2720 = ('DiagCode_2720', 'sum'),
totalDiagCode_2449 = ('DiagCode_2449', 'sum'),
totalDiagCode_4280 = ('DiagCode_4280', 'sum'),
totalDiagCode_53081 = ('DiagCode_53081', 'sum'),
totalDiagCode_41401 = ('DiagCode_41401', 'sum'),
totalDiagCode_496 = ('DiagCode_496', 'sum'),
totalDiagCode_2589 = ('DiagCode_2859', 'sum'),
totalDiagCode_41400 = ('DiagCode_41400', 'sum'),
totalDiagCode_Other = ('DiagCode_Other', 'sum'),
totalProcCode_4019 = ('ProcCode_4019.0', 'sum'),
totalProcCode_9904 = ('ProcCode_9904.0', 'sum'),
totalProcCode_2724 = ('ProcCode_2724.0', 'sum'),
totalProcCode_8154 = ('ProcCode_8154.0', 'sum'),
totalProcCode_66 = ('ProcCode_66.0', 'sum'),
totalProcCode_3893 = ('ProcCode_3893.0', 'sum'),
totalProcCode_3995 = ('ProcCode_3995.0', 'sum'),
totalProcCode_4516 = ('ProcCode_4516.0', 'sum'),
totalProcCode_3722 = ('ProcCode_3722.0', 'sum'),
totalProcCode_8151 = ('ProcCode_8151.0', 'sum'),
totalProcCode_8872 = ('ProcCode_8872.0', 'sum'),
totalProcCode_9671 = ('ProcCode_9671.0', 'sum'),
totalProcCode_4513 = ('ProcCode_4513.0', 'sum'),
totalProcCode_5849 = ('ProcCode_5849.0', 'sum'),
totalProcCode_9390 = ('ProcCode_9390.0', 'sum'),
totalProcCode_Other = ('ProcCode_Other', 'sum'))
# Caculate aggregted fraction
for column in self.fraction_column_list:
new_colunm = column + 'Frac_'
agg_df[new_colunm] = (X.groupby('Provider').apply(lambda x: (x[column] == 1).sum()/x[column].count())).values
# Total Claims per provider, unique benes per providers, claim/unique benes ratio
agg_df['ClaimNumbers'] = (X.groupby('Provider')[['ClaimID']].count()).values
agg_df['UniqBeneCount'] = (X.groupby('Provider')[['BeneID']].nunique()).values
agg_df['ClaimCountsperPatient'] = agg_df['ClaimNumbers'] / agg_df['UniqBeneCount']
agg_df['UniqATPhysCount'] = (X.groupby('Provider')[['AttendingPhysician']].nunique()).values
agg_df['ClmsperATPhysn'] = agg_df['ClaimNumbers'] / agg_df['UniqATPhysCount']
agg_df['UniqOPPhysCount'] = (X.groupby('Provider')[['OperatingPhysician']].nunique()).values
agg_df['ClmsperOPPhysn'] = agg_df['ClaimNumbers'] / agg_df['UniqOPPhysCount']
agg_df['UniqOTPhysCount'] = (X.groupby('Provider')[['OtherPhysician']].nunique()).values
agg_df['ClmsperOTPhysn'] = agg_df['ClaimNumbers'] / agg_df['UniqOTPhysCount']
# State, county count
agg_df['UniqStateCount'] = (X.groupby('Provider')[['State']].nunique()).values
return agg_df
def fit_transform(self, X, y=None):
self.fit(X)
return self.transform(X)
#######################################################################################
####################------All Steps Pipelines------####################################
#######################################################################################
steps = [('claim_period_transform', DateTransform(start='ClaimStartDt', end='ClaimEndDt', newColumn='ClaimPeriod')),
('hospital_period_transform', DateTransform(start='AdmissionDt', end='DischargeDt', newColumn='HospitalDays')),
('age_transform', AgeTransform(dob='DOB', dod='DOD', ageColumn='Age', deceasedColumn='Deceased')),
('diag_code_count', CodeCountTransform(colunmsToCount = diagnosis_code_columns, newColumn='DiagCodeCounts')),
('proc_code_count', CodeCountTransform(colunmsToCount = procedure_code_columns, newColumn='ProcCodeCounts')),
('diag_code_frequency_group', CodeFrequencyGroupTransform(code_columns=diagnosis_code_columns, new_columns_prefix='ClmDiag',
high=10000, medium_high=5000, medium=800, low=500)),
('proc_code_frequency_group', CodeFrequencyGroupTransform(code_columns=procedure_code_columns, new_columns_prefix='ClmProc',
high=500, medium_high=100, medium=10, low=5)),
('diag_code_top15_onehot', Top15OneHotTransform(column_list=diagnosis_code_columns, top_15_codes=train_top_15_diag_codes,
new_column_prefix='DiagCode_')),
('proc_code_top15_onehot', Top15OneHotTransform(column_list=procedure_code_columns, top_15_codes=train_top_15_proc_codes,
new_column_prefix='ProcCode_')),
('aggregation', ProviderLevelAggregateTransform(fraction_column_list=fraction_column_list))]
pipe = Pipeline(steps)
test_output = pipe.fit_transform(merge_provider_df)
print(test_output)