Research on Cargo Volume Prediction Based on SSA-BiLSTM-LGBM Fusion Models

In this paper, for the problem of cargo volume prediction in logistics network sorting centers, we constructed a SSA-BiLSTM-LGBM fusion cargo volume prediction model, and derived the daily and hourly cargo volume prediction results of 57 sorting centers in mid-December. First, the cargo volume data of 57 sorting centers were preprocessed; second, the sorting centers were divided into three major categories using K-means cluster analysis based on SAA optimization, and the Integrated Complementary Collective Empirical Modal Decomposition (ICCEMAN) was utilized to further analyze the characteristics of the cargo volume data. At the same time, the SSA-BiLSTM-LGBM fusion cargo volume prediction model was constructed to predict the sorting centers. The shipment volume is predicted, and the daily and hourly shipment volume of 57 sorting centers in December is obtained. In the case of changes in transportation routes, an exogenous variable (Route Influence Index) was constructed and integrated into the SSA-BiLSTM-LGBM fusion prediction model to obtain the daily, hourly shipment volume of the sorting centers in December. Finally, the ability of the constructed model to fit the data was measured, and the analysis concluded that the combination of multiple models provided better prediction results than a single model.