Optimizing Operational Efficiency in Logistics: a case study using Machine Learning Approaches

This thesis explores the optimization of outbound logistics operations through data-driven analysis and predictive modeling. Focusing on a real-world case study within a distribution center, the study investigates the relationship between workforce allocation, error occurrence, and productivity levels. Historical performance data were analyzed to identify high-efficiency patterns, which served as the foundation for developing predictive models aimed at simulating optimal resource distribution and anticipating future operational trends. A combination of Random Forest and XGBoost models was used to forecast production volumes across key activities, including automated sorting, manual sorting, and box closure. In contrast, a Gradient Boosting model was employed to predict the occurrence of outbound errors such as disruptions, improper packaging and incorrect sorting. The thesis assesses the economic impact of these operational simulations on overall revenue, highlighting how inefficiencies, even when minimal, can lead to substantial financial consequences. The proposed models provide actionable insights to enhance productivity, reduce error rates, and support strategic decision-making in logistics environments.