Research on emergency parcel mobilization in e-commerce logistics network based on BP neural network and Bayesian network prediction model

Logistics sites consist of receiving warehouses, sorting centers, business offices, etc. The e-commerce logistics network consists of logistics sites and transport routes connecting logistics sites. The volume of goods will be affected by the fluctuating situation of orders placed on holidays, force majeure disasters and other emergencies, which will lead to the shutdown of logistics venues or temporary inability to deliver goods, which will then need to be transmitted to other venues. Predicting the number of parcels in each logistics site and line can improve the operational efficiency and design an effective logistics network, which is of great significance for guaranteeing the normal transportation of the logistics network. The prediction model of the time series of the cargo volume of the line based on BP neural network is established. This paper visualizes the data of the daily cargo volume of each line during the period from 2023-01-01 to 2023-01-31, predict the data by using BP neural network, and output the prediction results for 31 days, taking the three lines DC14→DC10, DC20→DC35, DC25→DC62 as examples, and give the training diagram, training state, regression diagram, and the line simulation and prediction Results. The Bayesian network model is a directed acyclic graph (DAG), which consists of nodes representing variables and directed edges connecting these nodes. The nodes represent random variables and the directed edges between the nodes represent the interrelationships between the nodes, with conditional probabilities indicating the strength of the relationship between the variables. The absence of a parent node is represented as an input to that node using a priori probabilities. Multiple regression analysis was performed to test the correlation between the independent and dependent variables, using the volume of shipments as the dependent variable, and site 1 (sending courier) and site 2 (receiving courier) as the dependent variables. The average of all shipment volumes for each route for the years 2021 and 2022 was averaged as a proxy for the average shipment volume for each route. Multiple regression is performed. Finally robustness is tested.