Research on E-Commerce Logistics Path Optimization and Intelligent Scheduling System Based On Deep Reinforcement Learning

In the current big data era, the use of various algorithms and mathematical models to accurately predict the cargo flow of sorting centers has gradually evolved into an important trend in the development of the logistics industry. First of all, this paper visualizes and analyzes the daily and hourly freight volume of different sorting centers. Based on the visualization results, we found that the freight volume is roughly correlated with the feature values of lagged value, moving average, month, day, working day and trend. We functionalized the above selected features and built a random forest regression model to predict the daily and hourly freight volumes for the next 30 days. A scheduling model based on integer programming was then built based on the predictions and the number of available regular workers. The number of regular and temporary workers per shift at each sorting center for the next 30 days is derived. Finally, a heuristic algorithm was chosen to optimize the shift allocation. To ensure fairness in job scheduling and maximize employee efficiency, we chose to use a greedy algorithm - a local search algorithm - to solve the scheduling problem.