Explainable time-series forecasting with sampling-free SHAP for Transformers

TL;DR

SHAPformer is a novel Transformer-based time-series forecasting model that provides fast, accurate, and explainable predictions without sampling, enhancing transparency and trust in decision-making applications.

Contribution

It introduces a sampling-free, attention-based explanation method for time-series forecasting with Transformers, significantly improving explanation speed and fidelity.

Findings

Generates explanations in under one second, much faster than traditional methods.

Provides true-to-data explanations on synthetic datasets with ground truth.

Achieves competitive predictive performance on real-world electrical load data.

Abstract

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.