Exploration of Short Distance Freight Volume Prediction and Vehicle Scheduling Strategy Optimization

Abstract

In the logistics terminal operation scenario, accurate prediction of short distance transportation cargo volume is crucial for scientific vehicle scheduling, which is the key to improving logistics efficiency and reducing operating costs. This study utilizes data-driven predictive modeling and multi-objective programming scheduling modeling techniques to construct a complete analysis system, aiming to improve the accuracy of cargo volume prediction and optimize vehicle scheduling efficiency. Therefore, the research focuses on evaluating the robustness of the model in the face of prediction bias, providing a solid scientific basis for logistics enterprises' actual operational decisions. Based on LSTM model for time series prediction and gradient descent model for hyperparameter optimization, accurate prediction of time-based cargo volume is achieved. For the route of "Site 3- Station 83-0600", the coefficient of determination on the test set (R^2) reached 0.942, and the average absolute percentage error (MAPE) was 0.0432. For the "Site 3-station 83-1400" model, the average coefficient of determination (R^2) reached 0.9273 and the average MAPE was 0.1419, indicating good prediction accuracy and generalization ability of the model. Finally, the model decomposes the predicted results into a 10 minute granularity, providing data support for fine-grained regulation. The analysis and verification of the prediction results confirmed the accuracy of the prediction.