Title:Hierarchical Neural Additive Models for Interpretable Demand Forecasts

Abstract:Demand forecasts are the crucial basis for numerous business decisions, ranging from inventory management to strategic facility planning. While machine learning (ML) approaches offer accuracy gains, their interpretability and acceptance are notoriously lacking. Addressing this dilemma, we introduce Hierarchical Neural Additive Models for time series (HNAM). HNAM expands upon Neural Additive Models (NAM) by introducing a time-series specific additive model with a level and interacting covariate components.
Covariate interactions are only allowed according to a user-specified interaction hierarchy. For example, weekday effects may be estimated independently of other covariates, whereas a holiday effect may depend on the weekday and an additional promotion may depend on both former covariates that are lower in the interaction hierarchy.
Thereby, HNAM yields an intuitive forecasting interface in which analysts can observe the contribution for each known covariate. We evaluate the proposed approach and benchmark its performance against other state-of-the-art machine learning and statistical models extensively on real-world retail data. The results reveal that HNAM offers competitive prediction performance whilst providing plausible explanations.