AI-Driven Risk Management in Logistics: Predicting and Preventing Lost Orders Using Machine Learning

This paper was presented at the DSC Next Conference. The logistics sector plays a critical role in national economies, where the safe and timely transportation of orders is essential to minimizing loss and damage. This study proposes an AI-driven risk management system designed to reduce the percentage and overall number of lost orders within logistics operations. The proposed approach applies advanced machine learning techniques tailored to different transportation types, including pure transportation, collection, mainline transportation, and other logistics processes, each supported by distinct feature sets. The study evaluates the performance of CatBoost, LightGBM, and XGBoost models to predict whether an order is at risk of being lost. Following the prediction phase, a decision algorithm is applied to classify orders for effective risk management. The system achieves an accuracy of 92.12% for orders with no risk of loss and 45.79% for orders identified as at risk. The results demonstrate that the AI-driven approach significantly improves logistics risk management, leading to a reduction in lost orders and increased customer satisfaction.