LongxingTan/Time-series-prediction

tfts: Time Series Deep Learning Models in TensorFlow

What it solves

TFTS (TensorFlow Time Series) provides a unified, easy-to-use interface for implementing time series tasks such as prediction, classification, and anomaly detection. It simplifies the process of applying both classical and state-of-the-art deep learning models to time series data, reducing the boilerplate code required to set up training and evaluation pipelines in TensorFlow or Keras.

How it works

The library uses an AutoModel and AutoConfig system to allow users to quickly switch between different model architectures. It supports a wide range of architectures including RNNs, Transformers, Informer, Autoformer, and N-beats. Users can provide data in various formats (NumPy arrays, tf.data.Dataset, or tf.keras.utils.Sequence) and use the KerasTrainer to handle the training loop and visualization of results.

Who it’s for

This tool is designed for researchers, industry professionals, and competition participants who need to implement and train time series models quickly using the TensorFlow ecosystem.

Highlights

  • Broad Model Support: Includes SOTA models like Informer, Autoformer, TCN, BERT, and Temporal Fusion Transformer (TFT).
  • Flexible Data Input: Supports multiple data formats including NumPy arrays and TensorFlow datasets.
  • Customizability: Allows users to build custom models by using TFTS backbones with custom embedding layers or head layers for specific tasks.
  • Proven Performance: Used in winning entries for competitions like KDD Cup 2022 and Tianchi-ENSO prediction.

Related

  • Project
    AIStream-Peelout/flow-forecast

    An open-source deep learning framework for time series forecasting, classification, and anomaly detection featuring a wide array of state-of-the-art models.

  • Project
    tensorflow/tfjs

    A hardware-accelerated JavaScript library for training and deploying machine learning models in the browser and Node.js.

  • Project
    facebookresearch/Kats

    A comprehensive toolkit for time series analysis that provides a unified framework for forecasting, anomaly detection, and feature extraction.

  • Project
    tensorflow/models

    A collection of state-of-the-art model implementations and best practices for TensorFlow users to accelerate research and product development.

  • Project
    TensorSpeech/TensorFlowASR

    A TensorFlow-based framework for Automatic Speech Recognition that implements various ASR architectures and supports TFLite conversion for efficient deployment.