rapids/src/features/standardization/main.py

import pandas as pd
import numpy as np
from sklearn.preprocessing import StandardScaler

import sys

sensor_data_files = dict(snakemake.input)

provider = snakemake.params["provider"]
provider_key = snakemake.params["provider_key"]
sensor_key = snakemake.params["sensor_key"]

pd.set_option('display.max_columns', None)

if provider_key == "cr":
    sys.path.append('/rapids/src/features/')
    from cr_features_helper_methods import extract_second_order_features

    provider_main = snakemake.params["provider_main"]
    prefix = sensor_key + "_" + provider_key + "_"

    windows_features_data = pd.read_csv(sensor_data_files["windows_features_data"])
    excluded_columns = ['local_segment', 'local_segment_label', 'local_segment_start_datetime', 'local_segment_end_datetime', prefix + "level_1"]
    
    if windows_features_data.empty:
        windows_features_data.to_csv(snakemake.output[1], index=False)
        windows_features_data.to_csv(snakemake.output[0], index=False)
    else:
        windows_features_data.loc[:, ~windows_features_data.columns.isin(excluded_columns)] = StandardScaler().fit_transform(windows_features_data.loc[:, ~windows_features_data.columns.isin(excluded_columns)])

        windows_features_data.to_csv(snakemake.output[1], index=False)

        if provider_main["WINDOWS"]["COMPUTE"] and "SECOND_ORDER_FEATURES" in provider_main["WINDOWS"]:
            so_features_names = provider_main["WINDOWS"]["SECOND_ORDER_FEATURES"]
            windows_so_features_data = extract_second_order_features(windows_features_data, so_features_names, prefix)
            windows_so_features_data.to_csv(snakemake.output[0], index=False)
        else:
            pd.DataFrame().to_csv(snakemake.output[0], index=False)

else: 
    for sensor_features in sensor_data_files["sensor_features"]:
        if "/" + sensor_key + ".csv" in sensor_features:
            sensor_data = pd.read_csv(sensor_features)
            excluded_columns = ['local_segment', 'local_segment_label', 'local_segment_start_datetime', 'local_segment_end_datetime']

            if not sensor_data.empty:
                sensor_data.loc[:, ~sensor_data.columns.isin(excluded_columns)] = StandardScaler().fit_transform(sensor_data.loc[:, ~sensor_data.columns.isin(excluded_columns)])
            
            sensor_data.to_csv(snakemake.output[0], index=False)
            break