Graph Convolutional Network Based on Multidimensional Feature Fusion

• 1、School of Economics and Management, Beijing University of Posts and Telecommunications, Beijing 100876

Abstract：In order to effectively solve the problem of incomplete use of business scenario information in algorithms based on collaborative filtering, a multi-dimensional graph based graph neural network architecture is proposed. By utilizing the multi-dimensional graph generation strategy proposed in the study, the model can receive multiple types of data in the scene in the form of a multi-dimensional graph. At the same time, convolution is performed on each dimension through graph neural networks to obtain node representations on local dimensions. The feature fusion strategy is used to achieve the fusion of multi-dimensional features. Combining industrial datasets, experiments were conducted on the model to demonstrate the superiority of the model and algorithm.

Keywords： information management and information systems collaborative filtering graph neural network multidimensional graph feature fusion recommendation system