FEATURES["future_covariates_final"] = []
for col in FEATURES["future_covariates"]:
    new_features = data_preprocess[col].to_frame().copy()
    # Lag Features
    new_features[col+"_L0D"] = new_features[col].shift(0)
    new_features[col+"_L1D"] = new_features[col].shift(1)
    new_features[col+"_L2D"] = new_features[col].shift(2)

    # Rolling Features (No shift needed for future vars)
    new_features[col+"_RMean14D"] = new_features[col].rolling('14D').mean()
    new_features[col+"_RMin14D"] = new_features[col].rolling('14D').min()

    # Expanding Window (No shift needed for future vars)
    new_features[col+"_EMean14D"] = new_features[col].expanding(min_periods=14).mean()
    new_features[col+"_EMin14D"] = new_features[col].expanding(min_periods=14).min()
    
    FEATURES["future_covariates_final"].extend([col+"_L0D", col+"_L1D", col+"_L2D", col+"_RMean14D", col+"_RMin14D", col+"_EMean14D", col+"_EMin14D"])
    new_features = new_features.drop(columns=col)
    data_preprocess = pd.concat([data_preprocess, new_features], axis=1)

assert len(data_preprocess.loc[:, FEATURES["future_covariates_final"]].columns) == len(FEATURES["future_covariates"])*7