energy_fault_detector.quick_fault_detection.configuration
- select_config(train_data, normal_index, status_label_confidence_percentage, features_to_exclude, angles, automatic_optimization)
Selects a suitable config for the given data. The config is determined based on the data dimension and a PCA. If automatic optimization is True an optuna hyperparameter optimization is done for the autoencoder.
- Parameters:
train_data (pd.DataFrame) – Dataframe containing numerical values
normal_index (Union[pd.Series, None]) – Contains boolean information about which rows of train_data are normal and which contain anomalous behavior.
status_label_confidence_percentage (Optional float) – Determines the quantile for quantile threshold method.
features_to_exclude (Union[List[str], None]) – List of column names of train_data which should be ignored.
angles (Union[List[str], None]) – List of column names of angle features in train_data which need specialized preprocessing.
automatic_optimization (bool) – If True an optuna hyperparameter optimization is done for the autoencoder.
- Returns:
(optimized) config object for the AnomalyDetector.
- Return type:
- update_autoencoder_config(config, autoencoder_params)
Updates autoencoder parameters in the config with the new (optimized) parameters.
- update_preprocessor_config(config, features_to_exclude, angles)
Updates data preprocessor parameters in the config with user given metadata for features.
- Parameters:
- Returns:
Updated config object.
- Return type:
- update_threshold_config(config, quantile)
Updates threshold parameters in the config.