def get_patient_timeseries_label(no,label_name):
global PREP_OUTPUT_DIR, SAMPLE_PATIENT_PATH
# syntax checking existence for directory
PREP_OUTPUT_DIR = check_directory(PREP_OUTPUT_DIR)
output_path = PREP_OUTPUT_DIR + SAMPLE_PATIENT_PATH
x = pd.HDFStore(output_path,mode='r').select('data/{}'.format(label_name))
result_series = pd.Series(index=get_timeseries_column())
for _,x_date,x_label in x[x.no==no].values:
result_series[x_date]=x_label
return result_series
def set_age_dummies():
# 나이대를 설정해주는 함수
# main config파일에 어떻게 나이대를 설정할 것인지 나타나있음（AGE_BREAK_POINTS)
global demographic_output_path, AGE_BREAK_POINTS, AGE_LABELS
PREP_OUTPUT_DIR = check_directory(PREP_OUTPUT_DIR)
demographic_output_path = PREP_OUTPUT_DIR + DEMOGRAPHIC_OUTPUT_PATH
store_demo = pd.HDFStore(demographic_output_path, mode='r')
demo_except_age = store_demo.select('data/original', columns=['no', 'sex'])
demo_age = store_demo.select('data/original', columns=['age'])
cat_demo_age = pd.cut(demo_age.age, AGE_BREAK_POINTS, labels=AGE_LABELS)
cat_demo_age = cat_demo_age.cat.add_categories(['not known'])
cat_demo_age[cat_demo_age.isnull()] = 'not known'
cat_demo_age = pd.get_dummies(cat_demo_age)
_df = pd.concat([demo_except_age, cat_demo_age], axis=1)
store_demo.close()
_df.to_hdf(demographic_output_path,
'data/dummy',
format='table',
data_columns=True,
mode='a')
# -*- coding: utf-8 -*-
import sys, os, re
os_path = os.path.abspath('./') ; find_path = re.compile('emr_hypernatremia')
BASE_PATH = os_path[:find_path.search(os_path).span()[1]]
sys.path.append(BASE_PATH)
from generator.config import *
from generator.construct_common import check_directory, save_to_hdf5, get_timeseries_column, get_time_interval
# output path setting
global PREP_OUTPUT_DIR, PRESCRIBE_OUTPUT_PATH
PREP_OUTPUT_DIR= check_directory(PREP_OUTPUT_DIR)
prescribe_output_path = PREP_OUTPUT_DIR + PRESCRIBE_OUTPUT_PATH
def set_prescribe_row():
'''
약품코드를row_index_name으로나열
OFFSET_PRESICRIBE_COUNTS 기준에 따라서，drop 할 row을 결정
drop하고 남은 row를 metadata/usecol에 저장
'''
global OFFSET_PRESCRIBE_COUNTS, prescribe_output_path
store_pres = pd.HDFStore(prescribe_output_path,mode='r')
if not '/data' in store_pres.keys():
raise ValueError("There is no data in prescribe data")
pres_df = store_pres.select('data')
value_counts_diag = pres_df['medi_code'].value_counts()
total_pres = pres_df['medi_code'].sum()
min_value_counts = int(total_pres * OFFSET_PRESCRIBE_COUNTS)
if __name__ == "__main__":
global PREP_OUTPUT_DIR, SAMPLE_PATIENT_PATH
#argument
args = _set_parser()
label_name = args.label
label_list = list(range(1, int(args.label_range)))
core_num = int(args.core_num)
chunk_size = int(args.chunk_size)
time_length = int(args.time_length)
gap_length = int(args.gap_length)
target_length = int(args.target_length)
offset_min_counts = int(args.offset_min_counts)
offset_max_counts = int(args.offset_max_counts)
o_path = check_directory(args.path)
train_path = o_path + 'train/'
train_path = check_directory(train_path)
test_path = o_path + 'test/'
test_path = check_directory(test_path)
validation_path = o_path + 'validation/'
validation_path = check_directory(validation_path)
PREP_OUTPUT_DIR = check_directory(PREP_OUTPUT_DIR)
output_path = PREP_OUTPUT_DIR + SAMPLE_PATIENT_PATH
write_metadata_README(o_path, label_name, time_length, gap_length,
target_length, offset_min_counts, offset_max_counts)
sample_store = pd.HDFStore(output_path, mode='r')