A heterogeneous E-commerce user alignment model based on data enhancement and data representation

Abstract

Identifying users’ accounts on different platforms is important for the management and marketing of e-commerce platforms. To address the problem of sparse consumer behavior data on e-commerce platforms, this paper designs Double Generative Adversarial Network (Double-GAN) to compensate the data. This method uses not only the data of the one platform , but also of heterogeneous e-commerce platforms for alternate iteration compensation. For the complexity of “user-behavior-commodity” data feature space of e-commerce bookstore platform. Considering Consider the advantages of model JUST in capturing structural and semantic information of multiple types of nodes. This paper proposes a new method UBC2vec to integrate user’s interest in commodities and change the random walk strategy to represent the overall information of data feature space of heterogeneous e-commerce platform. Finally, for the high computational complexity of the user alignment algorithm, this paper constructs a “user-commodity” bipartite graph to delineate the roles of users and reduce the complexity. Experiments show that the model can effectively alleviate the data sparsity problem and reduce the computational complexity of the algorithm. Above the real data set, it can effectively match users’ accounts in different bookstore platforms.