Reproduce “RuleExplorer: A Scalable Matrix Visualization for Understanding Tree Ensemble Classifiers” on OpenResearch

Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.

Universal app for Apple silicon and Intel. Requires macOS 11 or later.

About this paper

Zhen Li, Weikai Yang, Jun Yuan, Jing Wu, Changjian Chen, Yao Ming, Fan Yang, Hui Zhang and 1 more

RuleExplorer is a visual analysis system enhancing the interpretability of tree ensemble classifiers by employing an anomaly-biased model reduction with a matrix-based hierarchical visualization. This approach helps users understand both common and anomalous decision rules within large rule sets, leading to improved model diagnosis and accuracy.

Key insights

  • Comprehensive model interpretation requires preserving both common and rare (anomalous) rules, as anomalies can highlight critical model flaws or unexpected behaviors.
  • Dynamic, user-driven hierarchical exploration of rules, rather than static reduction, is essential for experts to effectively navigate and analyze large and complex rule sets according to their specific analytical needs.
  • A visual analytics approach that quantifies rule anomaly and provides rich interactive visualizations allows experts to effectively identify data labeling errors, spurious correlations, and redundant features within models.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.