PHEATPRUNER: Interpretable Data-centric Feature Selection for Multivariate Time Series Classification through Persistent Homology

2025-04-25

AI Summary

Addressing the challenge of balancing performance and interpretability in multivariate time series classification, this paper proposes a data-driven topological feature selection method. Our approach innovatively integrates persistent homology with sheaf theory to construct an unsupervised, bias-free pruning framework—requiring neither labeled supervision nor posterior probabilities—to automatically reduce high-dimensional variables and generate structured interpretability vectors. Evaluated on the UEA benchmark and a real-world bovine mastitis dataset, our method achieves an average pruning rate of 45%, consistently improves (or at least maintains) classification accuracy, and incurs no additional computational overhead. It is compatible with mainstream classifiers—including Random Forest and XGBoost—demonstrating strong generalizability, intrinsic interpretability, and practical deployability.

Technology Category

• Machine Learning: Dimensionality Reduction/Feature Selection
• Data Mining & Knowledge Management: Data Compression
• Computer Vision: Interpretability, Explainability, and Transparency

Application Category

• Web Mining and Content Analysis: Robustness and generalizability of Web mining methods
• Graph Algorithms and Modeling for the Web: Algorithms and analysis for heterogeneous, signed, attributed, multi-relational, temporal, higher-order, and annotated Web-related graphs
• User Modeling, Personalization and Recommendation: Explainable and interpretable methods for personalization

Abstract

Balancing performance and interpretability in multivariate time series classification is a significant challenge due to data complexity and high dimensionality. This paper introduces PHeatPruner, a method integrating persistent homology and sheaf theory to address these challenges. Persistent homology facilitates the pruning of up to 45% of the applied variables while maintaining or enhancing the accuracy of models such as Random Forest, CatBoost, XGBoost, and LightGBM, all without depending on posterior probabilities or supervised optimization algorithms. Concurrently, sheaf theory contributes explanatory vectors that provide deeper insights into the data's structural nuances. The approach was validated using the UEA Archive and a mastitis detection dataset for dairy cows. The results demonstrate that PHeatPruner effectively preserves model accuracy. Furthermore, our results highlight PHeatPruner's key features, i.e. simplifying complex data and offering actionable insights without increasing processing time or complexity. This method bridges the gap between complexity reduction and interpretability, suggesting promising applications in various fields.

Problem

• Balancing performance and interpretability in multivariate time series classification
• Pruning high-dimensional data without sacrificing model accuracy
• Providing interpretable insights into complex data structure

Innovation

• Uses persistent homology for feature selection
• Integrates sheaf theory for explanatory insights
• Maintains accuracy while pruning 45% variables