Explainable Subgradient Tree Boosting for Prescriptive Analytics in Operations Management

Motivated by the overwhelming success of gradient boosting approaches in machine learning and driven by the need for explainable prescriptive analytics approaches in operations management (OM), we propose an explainable prescriptive analytics approach to solving complex OM problems: subgradient tree boosting (STB). The STB approach combines the well-known method of subgradient descent in function space with sample average approximation, and directly prescribes decisions from a problem-specific loss function, historical demand observations, and prescriptive features. The approach is inherently explainable and allows a decision-maker to derive detailed explanations for the prescribed decisions, such as a breakdown of individual features' impact that clarifies the underlying drivers of the prescriptions, which is increasingly useful in practice. We show how subgradients can be derived for many common OM problems, demonstrate STB's applicability to two real-world, complex capacity planning problems in the service industry, benchmark its performance against those of two prescriptive approaches—weighted sample average approximation and kernelized empirical risk minimization—and show how STB's prescriptions can be explained by estimating the impact of individual features. The results suggest that STB's performance is comparable to those of weighted sample average approximation and kernelized empirical risk minimization but also provides explainable prescriptions