This academic paper presents a method that combines explainable artificial intelligence (XAI) techniques with adaptive learning to improve energy consumption prediction models. The method focuses on handling shifts in data distribution. The process involves obtaining SHAP values to explain model predictions, clustering these values to identify patterns and outliers, and refining the model based on these characteristics. The approach was tested on a dataset of energy consumption records of buildings and showed improved predictive performance and interpretable model explanations.
Publication date: 7 Feb 2022
Project Page: https://arxiv.org/abs/2402.04982v1
Paper: https://arxiv.org/pdf/2402.04982
