让Transformer模型时间序列预测可解释，速度超快且无需采样。
Explainable time-series forecasting with sampling-free SHAP for Transformers
- 通过注意力机制操控特征子集，实现无采样的快速解释。
- 解释生成时间小于1秒，比传统方法快数个数量级。
- 适合需要实时可解释性的工业预测场景，如电力负荷分析。
时间序列预测在诸多领域中对规划和决策至关重要。可解释性是建立用户信任和满足透明性要求的关键。Shapley Additive Explanations (SHAP) 是一种流行的可解释AI框架，但其在时间序列上的高效实现仍缺乏，且常假设特征独立性来采样反事实数据。本文提出 SHAPformer，一个基于 Transformer 架构的准确、快速且无需采样的可解释时间序列预测模型。该模型通过操纵注意力机制，基于特征子集进行预测。SHAPformer 的解释生成时间不足一秒，相比 SHAP Permutation Explainer 快数个数量级。在具有真实解释的合成数据上，其生成的解释与数据一致；应用于真实世界电力负荷数据时，不仅预测性能具有竞争力，还揭示了局部与全局洞察，例如识别出历史负荷为关键预测因子，并发现圣诞期间模型行为显著不同。
原文摘要 · Abstract (English)
Time-series forecasts are essential for planning and decision-making in many domains. Explainability is key to building user trust and meeting transparency requirements. Shapley Additive Explanations (SHAP) is a popular explainable AI framework, but it lacks efficient implementations for time series and often assumes feature independence when sampling counterfactuals. We introduce SHAPformer, an accurate, fast and sampling-free explainable time-series forecasting model based on the Transformer architecture. It leverages attention manipulation to make predictions based on feature subsets. SHAPformer generates explanations in under one second, several orders of magnitude faster than the SHAP Permutation Explainer. On synthetic data with ground truth explanations, SHAPformer provides explanations that are true to the data. Applied to real-world electrical load data, it achieves competitive predictive performance and delivers meaningful local and global insights, such as identifying the past load as the key predictor and revealing a distinct model behavior during the Christmas period.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。