ProtoTSNet: Interpretable Multivariate Time Series Classification With Prototypical Parts

2025-11-04

AI Summary

Addressing the challenge of balancing high accuracy and interpretability in multivariate time series classification—particularly critical in high-stakes domains such as industrial monitoring and clinical decision-making—this paper proposes ProtoTS, an ante-hoc interpretable classification framework for multivariate time series. Methodologically, ProtoTS pioneers the adaptation of the prototype learning paradigm (ProtoPNet) to multivariate temporal data. It introduces a group-convolution-based encoder, pretrainable via autoencoding, that preserves channel-wise semantic meaning while quantifying feature importance. Coupled with prototype matching, it enables dynamic temporal pattern modeling and generates human-verifiable, prototype-level explanations. Evaluated on 30 UEA benchmark datasets, ProtoTS matches state-of-the-art black-box models in accuracy, substantially outperforms existing interpretable methods, and delivers expert-validated, instance-specific interpretations grounded in learned prototypes.

Technology Category

• Machine Learning: Time-Series/Data Streams
• Computer Vision: Interpretability, Explainability, and Transparency
• Humans and AI: Explainable AI (XAI) for Human Understanding

Application Category

• User Modeling, Personalization and Recommendation: Explainable and interpretable methods for personalization
• Security and Privacy: Data transparency and provenance
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications

Abstract

Time series data is one of the most popular data modalities in critical domains such as industry and medicine. The demand for algorithms that not only exhibit high accuracy but also offer interpretability is crucial in such fields, as decisions made there bear significant consequences. In this paper, we present ProtoTSNet, a novel approach to interpretable classification of multivariate time series data, through substantial enhancements to the ProtoPNet architecture. Our method is tailored to overcome the unique challenges of time series analysis, including capturing dynamic patterns and handling varying feature significance. Central to our innovation is a modified convolutional encoder utilizing group convolutions, pre-trainable as part of an autoencoder and designed to preserve and quantify feature importance. We evaluated our model on 30 multivariate time series datasets from the UEA archive, comparing our approach with existing explainable methods as well as non-explainable baselines. Through comprehensive evaluation and ablation studies, we demonstrate that our approach achieves the best performance among ante-hoc explainable methods while maintaining competitive performance with non-explainable and post-hoc explainable approaches, providing interpretable results accessible to domain experts.

Problem

• Interpretable classification for multivariate time series data
• Capturing dynamic patterns and varying feature significance
• Achieving competitive accuracy while providing domain-expert accessible explanations

Innovation

• Group convolutions in modified convolutional encoder
• Pre-trainable autoencoder preserving feature importance
• Interpretable classification capturing dynamic time series patterns