In the era of the booming e-commerce logistics industry, sorting centers are facing more stringent demands for cargo volume prediction and personnel scheduling. This research is centered around optimizing the human resource allocation within sorting centers. The approach is to build an effective cargo volume prediction model, with the dual goals of cutting down operational costs and enhancing sorting efficiency. We come up with the LSTM – Prophet model. This model merges the Prophet model, which is useful for time-series analysis, and the deep learning based LSTM network. Such a combination is designed to handle the complex and ever changing e-commerce logistics data. First, we carry out an in – depth analysis of the sorting center’s personnel structure, taking into account both regular employees and temporary workers. Then, based on the results of cargo volume predictions, we make rational arrangements for personnel shifts. We gathered and examined the daily shipment volumes from the past four months and the hourly throughput data of the previous 30 days, which were provided by 57 logistics centers. From the data analysis, we noticed that while most sorting centers shared common peak cargo volume points and general trends, there were still distinct differences in the trends of individual sorting centers. During the model-building process, we employed the Prophet model to identify trends, seasonal patterns, and holiday-related factors. Simultaneously, we utilized the LSTM network to address the long-term dependencies in the data. The LSTM – Prophet model can effectively fit the historical data through cell states and logical control units. Notably, it improves the prediction accuracy without the need to adjust the parameters of the Prophet model. The experimental results show that the LSTM-Prophet model performs remarkably well in predicting the daily and hourly – divided cargo volumes for the upcoming 30-day period. The model has strong explanatory power and high precision, as reflected by the adjusted R-squared values of 0.95005 and 0.94227 respectively, and the average relative errors of 0.08956 and 0.08918. In conclusion, the LSTM-Prophet model offers a novel shipment prediction tool for e-commerce logistics sorting centers. This tool is conducive to achieving more intelligent and automated operation management. It not only boosts the prediction accuracy but also provides guidance for sorting centers to allocate human resources rationally in various situations, thus reducing costs and increasing efficiency. In the future, we intend to further explore the application of this model in different scenarios and consider incorporating more influencing factors to further enhance its prediction accuracy and adaptability.