Towards Explainable Decision Support in Demand Forecasting: Global Feature Importance for Multi-Series Tree-Based Models

Abstract Global feature importance (GFI) methods are essential tools for interpreting machine learning models, yet their application in more complex forecasting tasks involving multiple time series can be challenging. This study focuses on tree-based ensemble models applied to multi-series product demand forecasting. To evaluate the different GFI methods, using controlled simulated datasets with controlled dependencies on lagged values and external demand drivers, we compare model-specific and model-agnostic global importance methods, including Shapley values, permutation importance, and tree-specific gain- and split-based importance. Our analysis focuses on uncovering pitfalls in applying these methods, including problems introduced by auto-correlation and feature scaling. This work contributes practical guidance for practitioners seeking to apply these methods in real-world forecasting scenarios and to leverage explainability methods for informed decision-making.