Supply-Chain Late-Delivery Risk Prediction and Resilience Analysis Using Machine Learning

Late deliveries cost penalties and customers, and a logistics planner can expedite only a limited number of orders. This project predicts at order time whether an order line of the real DataCo Smart Supply Chain dataset will ship late, compares a shipping-mode rule, a mode late-rate table and logistic regression with gradient-boosted trees (LightGBM and XGBoost with isotonic calibration) under a chronological split, explains the trees with SHAP and shows the leakage trap behind the very high accuracies published for this dataset. The risk is turned into decisions: cost-sensitive expediting, a Monte Carlo disruption study, a SimPy service-level simulation with time to recover, time-to-survive / time-to-recover linear programmes for a distribution-centre network and a sales-concentration (HHI) analysis. A web application scores orders, plans and approves expediting and runs what-if scenarios on five databases. A complete MATLAB version is included.

Category: Industrial Engineering

Industrial Engineering Project IE09 from ProTechSC Academy, with graphical abstract, video demonstration and project guidance.