def calc_month(df, month_type, build_type):
start_day = week_extract.get_min_month()
if build_type == TRAIN and month_type == MONTH1:
df_month = df[(df['date'] >= start_day)
& (df['date'] < timeOpt.add_months(start_day, 1))]
elif (build_type == TRAIN
and month_type == MONTH2) or (build_type == PREDICT
and month_type == MONTH1):
df_month = df[(df['date'] >= timeOpt.add_months(start_day, 1))
& (df['date'] < timeOpt.add_months(start_day, 2))]
else:
df_month = df[(df['date'] >= timeOpt.add_months(start_day, 2))
& (df['date'] < timeOpt.add_months(start_day, 3))]
group_month1 = day_extract.groupby_calc(df_month).apply(
month_extract.calc_month_data).reset_index(drop=True)
df = pd.DataFrame(group_month1) \
.rename(columns={'activeDay': 'activeDayMonth' + month_type,
'totalConnectTime': 'totalConnectTimeMonth' + month_type,
'dlTraffic': 'dlTrafficMonth' + month_type,
'ulTraffic': 'ulTrafficMonth' + month_type,
'totalDlUlTrafficMonthly': 'totalDlUlTrafficMonthly' + month_type,
'dlTrafficRatioMonthly': 'dlTrafficRatioMonthly' + month_type,
'totalDlUlTrafficPerday': 'totalDlUlTrafficPerdayMonth' + month_type,
'dlTrafficPerday': 'dlTrafficPerdayMonth' + month_type,
'ulTrafficPerday': 'ulTrafficPerdayMonth' + month_type,
'totalConnectTimePerday': 'totalConnectTimePerdayMonth' + month_type,
'MinRSRP': 'MinRSRPMonth' + month_type,
'MaxRSRP': 'MaxRSRPMonth' + month_type,
'AvgRSRP': 'AvgRSRPMonth' + month_type,
'StdRSRP': 'StdRSRPMonth' + month_type,
'MinSINR': 'MinSINRMonth' + month_type,
'MaxSINR': 'MaxSINRMonth' + month_type,
'AvgSINR': 'AvgSINRMonth' + month_type,
'StdSINR': 'StdSINRMonth' + month_type})
return df
def get_pre_month():
min_month = week_extract.get_min_month()
month_list = [(1, timeOpt.add_months(min_month,
1), timeOpt.add_months(min_month, 2)),
(2, timeOpt.add_months(min_month,
2), timeOpt.add_months(min_month, 3))]
return month_list
def get_post_df():
min_month = week_extract.get_min_month()
all_file = week_extract.get_file_by_range(timeOpt.add_months(min_month, 2),
timeOpt.add_months(min_month, 3))
df = pd.DataFrame(
columns=setting.parameter_json["post_eva_from_day_column_name"])
for f in all_file:
file_df = pd.read_csv(f, error_bad_lines=False, index_col=False)[
setting.parameter_json["post_eva_from_day_column_name"]]
df = df.append(file_df)
return df
def build_feature(df):
start_day = week_extract.get_min_month()
df_train = df[(df['date'] >= start_day)
& (df['date'] < timeOpt.add_months(start_day, 2))]
df_pre = df[(df['date'] >= timeOpt.add_months(start_day, 1))
& (df['date'] < timeOpt.add_months(start_day, 3))]
df_for_churn = set(df[(df['date'] >= timeOpt.add_months(start_day, 2))
& (df['date'] < timeOpt.add_months(start_day, 3))]
['esn'].astype('str').values)
compress.empty_folder(setting.model_path)
train_result_df = build(df_train, df_for_churn, TRAIN)
train_result_df.to_csv(os.path.join(setting.model_path, r"trainData.csv"),
index=False)
pre_result_df = build(df_pre, df_for_churn, PREDICT)
pre_result_df[(pre_result_df['churnLabel'] < 1)].to_csv(os.path.join(
setting.model_path, r"predictData.csv"),
index=False)
return 0
def run():
multiprocessing.freeze_support()
start_day = week_extract.get_min_month()
df = get_data_by_range(start_day, timeOpt.add_months(start_day, 3))
build_feature(df)