An Empirical Study of Explainable AI Techniques on Deep Learning Models For Time Series Tasks
This study addresses the problem of unverified outputs from Explainable AI (XAI) techniques for researchers and practitioners working with deep learning models on time series data, aiming to identify reliable attribution methods.
This paper proposes an empirical study and benchmark framework to apply and evaluate attribution methods, originally developed for image and text data, to deep learning models for time series tasks. It introduces a methodology to automatically evaluate and rank these attribution techniques on time series using perturbation methods to identify reliable approaches.
Decision explanations of machine learning black-box models are often generated by applying Explainable AI (XAI) techniques. However, many proposed XAI methods produce unverified outputs. Evaluation and verification are usually achieved with a visual interpretation by humans on individual images or text. In this preregistration, we propose an empirical study and benchmark framework to apply attribution methods for neural networks developed for images and text data on time series. We present a methodology to automatically evaluate and rank attribution techniques on time series using perturbation methods to identify reliable approaches.