Software & Resources

We develop and maintain open-source software for time-series analysis, from large feature-extraction toolkits to compact canonical feature sets, new analysis methods, and benchmark datasets. Everything is on our GitHub organization, where you will also find code for reproducing the results in our papers, and documented in detail in our time-series analysis tools guide.

Feature-extraction toolkits

Highly comparative toolkits that compute thousands of time-series features, or hundreds of pairwise interaction statistics, so that the best-performing methods for a problem can be found and interpreted systematically.

MATLAB

Extract thousands of time-series analysis features from a collection of time series.

GitHub Docs
Python

The Python successor to hctsa, and the most comprehensive time-series feature extraction package in Python. The active development path.

GitHub Docs
Python

A comprehensive library for computing pairwise interactions from multivariate time series.

GitHub Docs
R

Extraction, analysis, and visualisation of a range of open time-series features.

GitHub Docs

Canonical feature sets

Small, fast, interpretable sets of features distilled from the full hctsa library, for when you want strong performance without the computational cost.

CMATLABPythonRJulia

Evaluate 22 high-performing time-series features, coded efficiently in C and wrapped for python, with native versions for R and Julia.

GitHub Docs
CMATLABPython

Sixteen time-series features selected to characterise biologically informative dynamical patterns in mouse fMRI.

GitHub Docs

Methods

Implementations of new time-series analysis methods developed in the group. See also our feature-based information-theoretic method for detecting interpretable, long-timescale pairwise interactions.

PythonMATLAB

Rescaled Auto-Density, a noise-robust way to track the distance to criticality from time-series data.

GitHub Docs
Julia

Quantum-inspired time-series machine learning, modelling the joint distribution with matrix-product states.

GitHub Docs
Python

Python implementations of time-series dimension reduction methods from seven algorithm categories.

GitHub Docs

Datasets

Curated collections of time series with ground truth, for benchmarking and teaching.

Python

The 1000×1000 collection; 1000 simulated time series from 133 dynamical processes, each with ground truth, as a companion to Empirical1000.

GitHub Website

Resources

Guides and documentation we maintain for our tools, and for doing research more generally.

Time-series analysis tools guide Documentation hub for all of our software, with installation guides, tutorials and the papers behind each tool. catch22 feature documentation What each of the 22 catch22 features measures, with worked examples. Research resources Our advice for researchers, including doing a PhD, scientific writing, making figures, coding and computation, reproducible science, and organising your time. Pathways to a PhD in our group How to join the group as a research student, from first contact to scholarship applications. Visiting or moving to Sydney Practical notes for visitors and new arrivals. Our GitHub organization All of our software repositories, plus code and data for reproducing the results in our papers.

Our software is free for non-commercial use. For commercial licensing, or help applying these tools to your own data, see Engine Analytics.