eigen-entropy-tss coded by jkmackie

Python implementation of the Eigen-entropy based Time Series Signatures algorithm (EE - TSS) by jkmackie. To my knowledge, this is the first public implementation of EE - TSS.

The technique "achieves high recall rates with limited clinical datasets but also ensures the algorithm's feature generation is understandable, addressing a critical need for clinician-friendly tools." Also, it cuts the dimensions of high-dimensionality data by completely transforming it. The low dimension data--where the number of dimensions equals the number of scale factors--can then be fed into the classifier model.

The algorithm requires computational resources. It is coded in Python with Joblib parallel processing.

The binary classification dataset heartbeat is used to illustrate the algorithm. The classes are: 0=Normal and 1=Abnormal. Each heartbeat sample is a 405 observation time series. We must catch all abnormal heartbeats so recall is the key metric.

EE - TS results with heartbeat:

• For the original heartbeat data, the Ridge Classifier returns Recall = 98.64%
• EE - TSS transformed data and Ridge Classifier achieves Recall = 100%

$\color{#00ff00}{\textsf{TERMS OF USE: MIT No AI License}}$

Citations:

@code{
    author={jkmackie},
    title={eigen-entropy-tss},
    repo={https://github.com/jkmackie/eigen-entropy-tss},
    year={2025}
}

@article{
    author={Patharkar, A., Huang, J., Wu, T. et al.},
    title={Eigen‑entropy based time series signatures to support multivariate time series classification},
    year={2024},  
    journal={Nature Scientific Reports},
    url= {https://www.nature.com/articles/s41598-024-66953-7},
    doi={10.1038/s41598-024-66953-7}
}