Significance-based interpretable sequence clustering

https://www.sciencedirect.com/science/article/pii/S0020025525001045

Implementation Highlight

The significance-based clustering tree from the paper "Significance-based Interpretable Sequence Clustering" has been implemented in Matlab (SigISC.m), providing faster execution with the same results as the prior Python version (SigISC.py) in the co-author's repository.

View Our Method's Results

To view the final performance results as presented in the paper, simply run main.m in Matlab.

Datasets Used

The folder Sequence2BinaryData contains 14 real-world sequential datasets that have been profiled by specific patterns. These patterns have been mined using the "Sequential Pattern Discovery under Multiple Constraints" method (PatternMining.py).

Comparison Methods

This folder Comparison contains scripts for comparing various interpretable clustering methods, including IMM, SHA, and CUBT. The input data for these methods is identical to that used for constructing the decision tree with SigISC, as provided in the Sequence2BinaryData folder (.mat format in the main directory or .txt format in the Comparison directory). For the ISCT method, please refer to the official PyPI page.

How to Run the Comparison Methods

1. Running CUBT Methods:

   • Use the script seq_main_CUBT_two.R.
   • If the required libraries are not installed, you may need to install the corresponding R packages.

2. Running IMM and SHA Methods:

   • Use the script seq_main_IMM_SHA.py.
   • To avoid the hassle of package installation and compilation, I have provided a virtual environment (PyCharm venv).
   • Download the compressed file venv.7z from here.
   • Extract the file and import the environment as your Python interpreter (on Windows) to run the script.

Citation

If you find my code useful please consider citing:

@article{SigISC2025,
  title = {Significance-based interpretable sequence clustering},
  author = {Zengyou He and Lianyu Hu and Jinfeng He and Junjie Dong and Mudi Jiang and Xinying Liu},
  journal = {Information Sciences},
  volume = {704},
  pages = {121972},
  year = {2025},
  doi = {https://doi.org/10.1016/j.ins.2025.121972}
  }