Dr. Soundar Rajan Tirupatikumara, Thesis Supervisor Dr. Catherine Mary Harmonosky, Thesis Honors Advisor
Keywords:
Logistics Data Analytics Machine Learning
Abstract:
In today’s world, global manufacturing requires parts supplied from countries all over the world. Affordable labor cost in developing countries, coupled with the proximity to raw materials often lead to parts being manufactured thousands of miles away from the end customers. To focus on core competencies, Third Party Logistics (3PL) providers handle this supply chain aspect of companies. 3PL providers will estimate the delivery time and service level but the actual performance may deviate from their estimates. When a 3PL deviates, companies bear the unexpected supply chain costs since manufacturing resources operate under faulty estimates. We focus on parts and sub-assemblies that are procured from offshore suppliers. Using a data-driven approach, we forecast the shipment duration using travel-time information of previous shipments. Regression techniques and tree based models were used in this work. The model with highest R2 and lowest RMSE was chosen for the final application. We found that random forest model was best suited for the task. Given this model, a company can now forecast the expected shipment duration independent of 3PL providers’ estimates. The model was visualized in a business intelligence software to support plant managers in inventory tracking.