data = pd.read_csv('D:\\ParamsFolder\\ParamsFolder\\valid_params_without_noise(4s).csv')
------------------------------------------------------------------------------------------
wld_st_datetimes = [(datetime.fromtimestamp(timestamp/1000) + timedelta(hours=7)) for timestamp in data['start'].tolist()]
data['st_date'] = [x.date() for x in wld_st_datetimes]
data['st_time'] = [x.time() for x in wld_st_datetimes]
wld_datetimes = [(datetime.fromtimestamp(timestamp/1000) + timedelta(hours=7)) for timestamp in data['t'].tolist()]
data['t_date'] = [x.date() for x in wld_datetimes]
data['t_time'] = [x.time() for x in wld_datetimes]
--------------------------------------------------------------------------------------------
data_gene = pd.read_csv('D:\\ParamsFolder\\ParamsFolder\\IDs_hours_plusgen_v_And_03.csv', sep=';')
data_gene.drop(columns=['Unnamed: 1','ID,DataHours'], inplace=True)
--------------------------------------------------------------------------------------------
temp_data = data_gene[['ID', 'Gene 1 ACE', 'Gene 2']]
temp_data.columns = ['user', 'Gene 1', 'Gene 2']
merged_data = pd.merge(data, temp_data, on='user')
-------------------------------------------------------------------------------------------
%%time
new_df = pd.DataFrame()
users = merged_data['user'].unique()
for x in users:
    temp_df  = merged_data[merged_data['user'] == x]
    for i, row in temp_df.iterrows():
        indices = temp_df[['t_date', 't_time']].apply(lambda x: (x.t_date == row.t_date) & (x.t_time.hour == row.t_time.hour) &\
                                              (x.t_time.minute >= row.t_time.minute - 1) & \
                                              (x.t_time.minute <= row.t_time.minute + 1), axis=1).values
        agg_df = temp_df.iloc[indices].groupby('user').agg(['mean', 'std']).reset_index()
        new_df = pd.concat([new_df, agg_df], ignore_index=True)