Multi-Task Learning-Based Optimization for Cross-Regional Logistics Scheduling and Transportation Efficiency

Abstract

With the development of globalization and digitalization, improving the efficiency of cross-regional logistics scheduling has become a key issue. Traditional logistics optimization algorithms have limitations in complex multi-task scenarios. Multi-task learning, as a branch of deep learning, provides a new idea for solving this problem. This paper proposes a logistics scheduling optimization strategy and model based on multi-task learning. This study is evaluated based on a real data set containing 5,000 crossregional logistics order records and a simulated data set covering 300 different transportation scenarios. The real data set comes from the logistics business of 5 major logistics hub cities in China and their radiating areas within half a year, and the simulated data set is constructed by comprehensively considering factors such as terrain, traffic conditions, and order density in different regions. The model has achieved an order allocation accuracy of 92%, the path planning cost is reduced by 25% compared with the traditional method, and the transportation time prediction error is controlled within ±3 hours. Among them, the order allocation accuracy is calculated as the proportion of the number of correctly allocated orders to the total number of orders, the path planning cost is obtained by combining the actual transportation mileage with the unit mileage cost, and the transportation time prediction error is the average of the absolute value of the difference between the predicted time and the actual transportation time. The model has achieved an order allocation accuracy of 92%, the path planning cost is reduced by 25% compared with the traditional method, and the transportation time prediction error is controlled within ±3 hours. Among them, the order allocation accuracy is calculated as the ratio of the number of correctly allocated orders to the total number of orders. The path planning cost is obtained by combining the actual transportation mileage with the unit mileage cost. The transportation time prediction error is the average of the absolute value of the difference between the predicted time and the actual transportation time.