文章背景与核心概要
本文是一篇关于时间序列分类(TSC)中可解释人工智能(XAI)软件框架的系统性综述。随着时间序列数据在许多关键决策领域的广泛应用,确保模型透明度与可信度变得至关重要,然而现有的XAI研究大多处于碎片化状态,缺乏对可用软件框架及其评估实践的系统性理解。
通过对六个显式支持时间序列的软件框架进行评估,作者揭示了当前生态系统中的几个关键局限性:首先,尽管频域解释具有高度相关性,但仅有一种方法对其提供支持;其次,专门针对时间序列数据量身定制的评估指标仅有两个;最后,使用不同的框架时,相同的XAI方法可能会产生截然不同的解释结果。这项研究强调了建立统一且具备时间序列感知能力的XAI框架的紧迫需求,从而确保解释的忠实性、可复现性和稳健性。
Software Frameworks for Explainable AI in Time Series Classification: A Systematic Review

Summary
本系统性综述探讨了专为时间序列分类(TSC)设计的可解释人工智能(XAI)软件框架的现状。尽管时间序列数据被大量应用于决策关键型机器学习应用中,但其可解释性研究在很大程度上仍处于碎片化状态。
通过评估六个明确支持时间序列的软件框架,作者揭示了当前生态系统中的关键局限性: * 频域支持: 尽管频域解释高度相关,但仅有一种方法对其提供了支持。 * 评估指标: 仅有两个评估指标是专门针对时间序列数据量身定制的。 * 跨框架不一致性: 根据所使用的框架,相同的XAI方法可能会产生截然不同的解释。
本文强调,迫切需要统一且具备时间序列感知能力的XAI框架,以确保解释具有忠实性、可复现性和稳健性。
This systematic review explores the landscape of Explainable Artificial Intelligence (XAI) software frameworks designed specifically for Time Series Classification (TSC). While time series data is heavily utilized in decision-critical machine learning applications, research into its explainability has largely remained fragmented.
Evaluating six software frameworks that explicitly support time series, the authors reveal critical limitations in the current ecosystem: * Frequency-Domain Support: Only a single method supports frequency-domain explanations, despite their high relevance. * Evaluation Metrics: Merely two evaluation metrics have been tailored specifically for time series data. * Cross-Framework Inconsistency: Identical XAI methods can yield vastly different explanations depending on the framework used.
The paper highlights an urgent need for unified, time-series-aware XAI frameworks that guarantee faithful, reproducible, and robust explanations.
Article Metadata
| 字段 | 详情 |
|---|---|
| arXiv ID | arXiv:2608.21449 [cs.AI] |
| 作者 | Louis Peter, Nils Gumpfer, Jana Fischer, Christin Seifert, Jennifer Hannig |
| 提交时间 | 2026年8月19日 |
| 主学科 | 人工智能 (cs.AI) |
| ACM 分类 | I.2.6; I.5.1; D.2.13; A.1 |
| 备注 | 17页,5张图表,将发表于 ECML PKDD 2026, XKDD and Beyond Workshop, LNCS |
| DOI | 10.48550/arXiv.2608.21449 |
Article Metadata
Field Details arXiv ID arXiv:2608.21449 [cs.AI] Authors Louis Peter, Nils Gumpfer, Jana Fischer, Christin Seifert, Jennifer Hannig Submitted August 19, 2026 Primary Subject Artificial Intelligence ( cs.AI)ACM Classes I.2.6; I.5.1; D.2.13; A.1 Comments 17 pages, 5 figures, to be published in ECML PKDD 2026, XKDD and Beyond Workshop, LNCS DOI 10.48550/arXiv.2608.21449
Abstract
时间序列广泛存在于各种应用领域中,并在决策关键型场景中通过机器学习进行分析。时间序列分类(TSC)是最受广泛研究且最具相关性的任务之一。在此背景下,确保TSC模型的透明度与可信度已成为一项重要需求,这也推动了可解释人工智能(XAI)方法的使用。
尽管人们的兴趣日益浓厚,但关于TSC中XAI的研究仍然呈碎片化状态,并且仍然缺乏对用于生成解释的可用软件框架、其评估实践以及实际局限性的系统性理解。先前的工作主要集中于单个解释方法,而跨框架的一致性、时间序列特定的评估以及可复现性则很少受到关注。
在本综述中,我们分析了用于TSC中解释生成和评估的现有软件框架。我们在多个维度上对它们进行了比较,包括支持的XAI方法、评估指标、可用性、基准测试支持以及可复现性,从而提供了首个针对时间序列的框架综述,其中包含实现比较和频域支持分析。我们识别出了六个明确支持时间序列的框架,并揭示了其共同的局限性: 1. 尽管频域解释具有相关性,但只有一种方法对其提供支持。 2. 只有两个评估指标是专门针对时间序列开发的。 3. 跨框架使用相同的XAI方法可能会产生大相径庭的解释。
基于这些发现,我们讨论了开放性挑战并概述了未来研究的方向,突出了对统一的、具备时间序列感知能力的XAI框架的需求,这些框架能够实现忠实、可复现且具时间序列感知的解释。
Time series arise in a wide range of application domains and are analyzed using machine learning in decision-critical settings. Time series classification (TSC) is one of the most widely studied and relevant tasks. In this context, ensuring the transparency and trustworthiness of TSC models has become an important requirement, motivating the use of explainable artificial intelligence (XAI) methods.
Despite growing interest, research on XAI for TSC remains fragmented, and a systematic understanding of the available software frameworks for explanation generation, their evaluation practices, and practical limitations is still lacking. Prior work largely focused on individual explanation methods, while cross-framework consistency, time-series-specific evaluation, and reproducibility have received little attention.
In this survey, we analyze existing software frameworks for explanation generation and evaluation in TSC. We compare them along multiple dimensions, including supported XAI methods, evaluation metrics, usability, benchmarking support, and reproducibility, providing the first time-series-specific survey of frameworks with implementation comparisons and an analysis of frequency-domain support. We identify six frameworks that explicitly support time series and reveal common limitations: 1. Only one method supports frequency-domain explanations despite their relevance. 2. Only two evaluation metrics have been developed specifically for time series. 3. Identical XAI methods can yield substantially different explanations across frameworks.
Based on these findings, we discuss open challenges and outline directions for future research, highlighting the need for unified, time-series-specific XAI frameworks that enable faithful, reproducible, and time-series-aware explanations.
Full-Text & Access Links
- PDF: 查看 PDF
- HTML(实验性): HTML 版本
- TeX 源码: 下载源码
- 许可证: 知识共享署名 4.0 国际许可协议
Full-Text & Access Links
- PDF: View PDF
- HTML (Experimental): HTML Version
- TeX Source: Download Source
- License: Creative Commons Attribution 4.0 International