Bi-Level Cargo Volume Prediction Method Based on Machine Learning Approach

In recent years, with the rapid development of the e-commerce industry, logistics has become pivotal, and the sorting platform, as the core of logistics and distribution, has long been characterized by problems such as large fluctuations in task demand and complex personnel scheduling. This paper seeks to accurately predict the cargo volume of the sorting center through machine learning methods including exponential smoothing and Gaussian process regression. The dataset used in this paper is from the 14th MathorCup Mathematics Application Challenge 2024. time series and three times exponential smoothing methods are used to predict the daily cargo volume, the prediction accuracy reaches 89.7%, and the RMSE value is 1.836, the MAE value is 1.213. The prediction result is better than the traditional ARIMA model. In addition, the noise caused by many uncertainties and the problems caused by changes in transportation routes were considered. To solve this problem, a Gaussian process regression prediction is established, which is characterized by time series and daily cargo volume, and the prediction for the next 30 days is relatively accurate.