/time-series

Time-Series models for multivariate and multistep forecasting, regression, and classification

Primary LanguagePythonMIT LicenseMIT

Time-Series

time-series is a Python module for machine learning for time-series built on top of tensorflow and is distributed under the MIT license.

This repository was created as a companion repository for chapter 12, Multivariate Forecasting, of the book Machine Learning for Time-Series with Python.

Tensorflow implementations of Time-Series models including these:

  • Amazon DeepAR
  • Gaussian Processes
  • LSTM
  • TCN
  • Transformer, and
  • NBEATS

The time_series.dataset package, part of this library, includes utility functions for loading datasets.

Please see the example notebook for usage and training results.

Installation

pip install git+https://github.com/benman1/time-series.git

Contribute

Pull requests welcome!

List of Contributors

Contributions from various people have found their way into this repository. Thanks to everyone for their hard work!