Skip to content

TimeSeriesFeature

TimeSeriesFeature with support for lag features and temporal processing.

Constructor

__init__(self, name: str, feature_type: kdp.features.FeatureType = <FeatureType.TIME_SERIES: 10>, lag_config: dict = None, rolling_stats_config: dict = None, differencing_config: dict = None, moving_average_config: dict = None, wavelet_transform_config: dict = None, tsfresh_feature_config: dict = None, calendar_feature_config: dict = None, sequence_length: int = None, sort_by: str = None, sort_ascending: bool = True, group_by: str = None, dtype: tensorflow.python.framework.dtypes.DType = tf.float32, is_target: bool = False, exclude_from_input: bool = False, input_type: str = 'continuous', **kwargs) -> None

Initializes a TimeSeriesFeature instance.

Parameters- name (str): The name of the feature.

feature_type (FeatureType): The type of the feature.
lag_config (dict): Configuration for lag features. If None, no lag features will be created.
  • Example: {'lags': [1, 7, 14], 'drop_na': True} rolling_stats_config (dict): Configuration for rolling statistics.
  • Example: {'window_size': 7, 'statistics': ['mean', 'std']} differencing_config (dict): Configuration for differencing.
  • Example: {'order': 1} moving_average_config (dict): Configuration for moving averages.
  • Example: {'periods': [7, 14, 30]} wavelet_transform_config (dict): Configuration for wavelet transform.
  • Example: {'levels': 3, 'window_sizes': [4, 8, 16], 'flatten_output': True} tsfresh_feature_config (dict): Configuration for statistical feature extraction.
  • Example: {'features': ['mean', 'std', 'min', 'max'], 'normalize': True} calendar_feature_config (dict): Configuration for calendar features.
  • Example: {'features': ['month', 'day', 'day_of_week'], 'cyclic_encoding': True} sequence_length (int): Length of the sequence. If None, no sequence will be created. sort_by (str): Column name to sort the time series data by (typically a timestamp column). Required for proper time series ordering. sort_ascending (bool): Whether to sort in ascending order (True) or descending order (False). Default is True for chronological ordering. group_by (str): Optional column name to group time series data by. Useful for multiple series (e.g., data for different stores, customers, products, etc.) dtype (tf.DType): The data type of the feature. is_target (bool): Whether this feature is a target for prediction. exclude_from_input (bool): Whether to exclude this feature from the input. input_type (str): The input type of the feature (e.g., "continuous"). **kwargs: Additional keyword arguments for the feature.

add_preprocessor

add_preprocessor(self, preprocessor: kdp.layers_factory.PreprocessorLayerFactory | typing.Any) -> None

Adds a preprocessor to the feature.

Parameters- preprocessor (Union[PreprocessorLayerFactory, Any]): The preprocessor to add.

build_layers

build_layers(self, row_preserving: bool = True) -> list

Build the appropriate layers for this time series feature based on configuration.

Parameters- row_preserving: When True (the default) the layers keep every input

    row, padding the warm-up positions instead of dropping them.
    A preprocessing model lays features out side by side, so a layer
    that removes its feature's leading rows leaves that column
    shorter than every other one and the concatenation fails. Pass
    False only when driving the returned layers directly.

Returns

  • list: List of TensorFlow layers for time series preprocessing

from_string

from_string(type_str: str) -> 'FeatureType'

Converts a string to a FeatureType.

Parameters- type_str (str): The string representation of the feature type.

get_output_dim

get_output_dim(self) -> int

Calculate the output dimension of this feature after all transformations.

The layers built- by: meth:build_layers are applied in sequence, and each one that keeps its originals passes them through alongside its new columns. The widths therefore compose multiplicatively, not additively: two lags on a single column give 3 columns, and differencing that while keeping the originals gives 6, not 4.

Returns

  • int: The number of columns the layer stack produces.

to_dict

to_dict(self) -> dict

Convert the feature configuration to a dictionary.

Returns- dict: Dictionary representation of the feature

update_kwargs

update_kwargs(self, **kwargs) -> None

Updates the kwargs with new or modified parameters.

Parameters

**kwargs: The new or modified parameters.