Module: tf.contrib.timeseries

Stay organized with collections Save and categorize content based on your preferences.

View source on GitHub

A time series library in TensorFlow (TFTS).


saved_model_utils module: Convenience functions for working with time series saved_models.


class ARModel: Auto-regressive model, both linear and non-linear.

class ARRegressor: An Estimator for an (optionally non-linear) autoregressive model.

class CSVReader: Reads from a collection of CSV-formatted files.

class FilteringResults: Keys returned from evaluation/filtering.

class NumpyReader: A time series parser for feeding Numpy arrays to a TimeSeriesInputFn.

class OneShotPredictionHead: A time series head which exports a single stateless serving signature.

class RandomWindowInputFn: Wraps a TimeSeriesReader to create random batches of windows.

class StructuralEnsembleRegressor: An Estimator for structural time series models.

class TimeSeriesRegressor: An Estimator to fit and evaluate a time series model.

class TrainEvalFeatures: Feature names used during training and evaluation.

class WholeDatasetInputFn: Supports passing a full time series to a model for evaluation/inference.


predict_continuation_input_fn(...): An Estimator input_fn for running predict() after evaluate().