Robust Supply Chains with Gradient Boosted Trees

Pradeep K. Mahato, Apurva Narayan

Abstract: Supply chain networks often experience various internal and external events that lead to shipment failures. Despite advancements in various machine learning models, the problem of avoiding service level failures remains intricate and hard to solve. While multiple attempts have been made by various researchers to make supply chains resilient, this is still an open problem. Moreover, explainability in the field of machine learning is a challenging task that assists in decision formation along with transparency.We develop a machine learning pipeline with gradient boosted decision trees to mitigate service level failures in supply chains. Our framework is simple, easy to implement, and provides a promising result. It provides explainability to prevent service level failure in time sensitive supply chains such as food manufacturing. Our model can be used for rapid deployment with state-of-the-art prediction accuracy while establishing trust within the decision-makers.