b523b464d8cafe29e352981c1c6df941f205592a,nilmtk/metrics.py,,mean_normalized_error_power,#Any#Any#,109

Before Change

    """

    mne = {}
    numerator = {}
    denominator = {}

    for appliance in predicted_power:
        numerator[appliance] = np.sum(np.abs(predicted_power[appliance] -
                                             df_appliances_ground_truth[appliance].values))
        denominator[appliance] = np.sum(
            df_appliances_ground_truth[appliance].values)
        mne[appliance] = numerator[appliance] * 1.0 / denominator[appliance]
    return mne


def rms_error_power(predicted_power, df_appliances_ground_truth):

After Change

        pred_generator = meter.power_series(periods=sections)
        total_diff = 0
        sum_of_ground_truth_power = 0
        while True:
            try:
                pred_chunk = next(pred_generator)
            except StopIteration:
                break
            else:
                truth_generator = ground_truth_meter.power_series(
                    periods=[pred_chunk.timeframe], chunksize=1E9)
                truth_chunk = next(truth_generator)
                
                // TODO: do this resampling in the pipeline?
                truth_chunk = truth_chunk.resample(period_alias)
                pred_chunk = pred_chunk.resample(period_alias)

                diff = (pred_chunk.icol(0) - truth_chunk.icol(0)).dropna() 
                total_diff += sum(abs(diff))
                sum_of_ground_truth_power += truth_chunk.icol(0).dropna().sum()

        mne[meter.instance()] = total_diff / sum_of_ground_truth_power

    return pd.Series(mne)

Instances