Cargo Volume Prediction Based on K-means Clustering and Back Propagation Neural Network

Abstract

Volume forecasting for sorting centers in e-commerce logistics networks is important for optimizing resource scheduling and transportation efficiency. In this paper, a joint prediction method incorporating K-means clustering algorithm and Back Propagation neural network (BP neural network) is proposed for 57 sorting centers after route adjustment. Firstly, based on the historical transportation data, we construct the logistics network topology map and identify the key hub nodes, and then divide the sorting centers into 3 classes by K-means algorithm to reveal the characteristics of their cargo volume distribution. Then, combined with the information of future transportation route changes, the training and prediction sets of daily and hourly cargo volume are constructed, and the BP neural network model is used for multi-scale prediction. Finally, Analysis of Variance (ANOVA) further shows that the clustering results are statistically significant, indicating that there are significant differences in the shipment volumes of different categories of sorting centers. In addition, in order to test the accuracy of the BP neural network, this paper conducted mean square error and regression analysis to verify the reliability of the prediction results. The model provides a scientific basis for personnel scheduling and transportation optimization in sorting centers, and has high robustness and practical value.