Research on the Classification of E-commerce Network Consumers based on Improved LPA Algorithm

In-depth research on the types and characteristics of consumers in e-commerce networks can help e-commerce operators achieve precise marketing. Aiming at the shortcomings of the Label Propagation Algorithm (LPA) in the node update order and label propagation rules in consumer type identification, a consumer type classification algorithm based on the improved LPA algorithm is proposed in e-commerce networks. Due to the different influence of consumers in the network, the node shrinkage method is introduced to calculate the influence of nodes, and the node update order is determined according to the descending sequence of the influence of nodes. Due to the different trust degrees among consumers, when there are multiple labels with the most occurrence times, the Jaccard index is introduced to calculate the trust degree between the target node and the neighbor nodes carrying these candidate labels, and the node with the largest trust degree is selected to update the label. Experiments show that in the e-commerce network built based on consumer user behavior data, compared with the LPA algorithm, the average modularity Q value of the algorithm is increased by 0.0689, and the division result is better than that of the LPAm and NIBLPA algorithms, indicating that the algorithm is more Suitable for e-commerce networks, the type division results are more accurate.