Scaling Time Series Classification via XAI-Driven Data Reduction

2026-07-17

AI Summary

This work addresses the scalability bottleneck of Transformer-based models in time series classification, which stems from their high computational complexity. To overcome this limitation, the authors propose a novel dimensionality reduction approach grounded in explainable artificial intelligence (XAI). The method leverages a GPU-accelerated Hydra classifier to generate local attributions, which are aggregated into global feature importance scores. An automatic elbow-point heuristic adaptively selects the most relevant temporal features without requiring manual threshold tuning. Notably, this study pioneers the repurposing of XAI from a post-hoc interpretability tool into a mechanism for scalable data reduction. Evaluated on real-world datasets, the approach achieves 80%–90% data compression while preserving classification accuracy comparable to that obtained with full-length sequences, thereby enabling large models such as ConvTran to process datasets previously infeasible due to memory constraints.

Abstract

Explainable AI (XAI) for time series has seen significant algorithmic growth, but its utility in providing measurable performance gains for downstream tasks remains under-explored. This paper bridges this gap by introducing drXAI, a novel methodology that repurposes XAI attribution methods for effective data reduction in Time Series Classification (TSC). The core challenge in modern TSC is scalability; state-of-the-art models, such as Transformers, exhibit quadratic complexity relative to sequence length and linear complexity relative to the number of channels. This renders them computationally prohibitive for massive datasets. drXAI addresses this by using a fast, GPU-accelerated classifier (Hydra) to generate local attributions. We aggregate these into global feature importance scores and employ an automated elbow-cut heuristic to select the most salient features without requiring manual thresholds. We evaluate our approach on both synthetic and real-world univariate and multivariate datasets. On synthetic benchmarks, drXAI successfully recovers ground-truth features where traditional baselines fail. On real-world data, drXAI achieves between 80% and 90% data reduction while maintaining classification accuracy comparable to models trained on the full dataset. Most importantly, we show that drXAI allows resource-intensive models like ConvTran to scale to datasets that were previously inaccessible due to memory constraints. Our results show the benefits of using XAI not just for interpretability, but as a robust tool for feature selection and scalability in time series analysis. All our code and data are openly available.

Problem

• Time Series Classification
• Scalability
• Data Reduction
• Computational Complexity
• XAI

Innovation

• XAI-driven data reduction
• Time Series Classification
• feature selection
• scalability
• attribution methods