benfulcher/hctsa

Highly comparative time-series analysis

What it solves

It provides a framework for "highly comparative time-series analysis," allowing users to extract thousands of features from univariate time series to characterize their behavior and identify patterns that distinguish different classes of data.

How it works

The software operates as a Matlab package that processes collections of univariate time series through a specific pipeline: it formats the data, computes a massive set of features, normalizes the resulting feature matrix, and then uses clustering and visualization tools to analyze the relationships between the time series and the features themselves.

Who it’s for

Researchers and data scientists working with time-series data who need to perform automated phenotyping or comprehensive feature extraction for classification and analysis.

Highlights

  • Extracts thousands of time-series features automatically.
  • Includes tools for normalizing and clustering time-series data.
  • Generates low-dimensional representations of time-series data.
  • Identifies and interprets discriminating features between different classes.
  • Fits and evaluates multivariate classification models.
  • Supports parallel computing via the Matlab Parallel Computing Toolbox.

Related

  • Project
    facebookresearch/Kats

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

  • Project
    Nixtla/tsfeatures

    A Python implementation of the R tsfeatures package that calculates a wide variety of of summary features from time series data for analysis and meta-learning.

  • Project
    decisionintelligence/TFB

    TFB is an open-source library for the fair and comprehensive benchmarking of time series forecasting methods, providing a standardized pipeline to prevent evaluation biases.

  • Project
    Nixtla/statsforecast

    A high-performance library for univariate time series forecasting that provides fast, scalable implementations of statistical and econometric models.

  • Project
    LongxingTan/Time-series-prediction

    TFTS is a TensorFlow-based library that provides a unified interface for implementing state-of-the-art deep learning models for time series prediction, classification, and anomaly detection.