Optimizing Inventory Routing: A Decision-Focused Learning Approach using Neural Networks
MD Shafikul Islam, Azmine Toushik Wasi
TL;DR
The paper addresses the Inventory Routing Problem ($IRP$) under demand uncertainty and critiques the two-stage Predict-then-Optimize paradigm for suboptimal decisions. It proposes Decision-Focused Learning ($DFL$) to integrate demand prediction with routing optimization in an end-to-end framework, enabling backpropagation of decision loss through the optimization layer. Two differentiable strategies are explored: a regularized Quadratic Program surrogate via implicit differentiation of the KKT conditions and a log-barrier interior-point formulation to preserve feasibility. The experiments show that emphasizing predictive accuracy alone is insufficient for profitability and discuss practical scalability considerations for NP-hard $IRP$. The findings motivate a robust, decision-focused approach to inventory and routing under uncertainty with potential real-world impact on supply chain resilience.
Abstract
Inventory Routing Problem (IRP) is a crucial challenge in supply chain management as it involves optimizing efficient route selection while considering the uncertainty of inventory demand planning. To solve IRPs, usually a two-stage approach is employed, where demand is predicted using machine learning techniques first, and then an optimization algorithm is used to minimize routing costs. Our experiment shows machine learning models fall short of achieving perfect accuracy because inventory levels are influenced by the dynamic business environment, which, in turn, affects the optimization problem in the next stage, resulting in sub-optimal decisions. In this paper, we formulate and propose a decision-focused learning-based approach to solving real-world IRPs. This approach directly integrates inventory prediction and routing optimization within an end-to-end system potentially ensuring a robust supply chain strategy.