Research on Product Recommendation System of Online Mall Based on Graph Neural Network

In the current era of Internet and e-commerce, the research and application of product recommendation system in online mall has become an important field. With the rapid development of big data technology and the accumulation of user behavior data, the product recommendation system of online mall has also become the core of the recommendation system. At present, a variety of recommendation algorithms, including collaborative filtering, content-based recommendation and deep learning, have been widely studied and applied. Online mall product recommendation needs to have real-time, can quickly respond to user interaction and changes in commodity inventory. Therefore, the recommendation system needs to have efficient algorithm and technology, as well as reasonable system architecture to support real-time recommendation service. However, in the past, the product recommendation system in the online mall was often simple and limited. Simple methods such as popularity-based, content-based and collaborative filtering are mainly used for recommendation, but there are some problems such as insufficient personalization and inaccurate characterization of user characteristics. Traditional recommendation systems tend to recommend products to users that they already like, but this can lead to problems of information filtering and user interest limitation. In addition, the traditional recommendation method is still based on the similarity of product content to recommend products with similar characteristics to users through the analysis of product text description, label and other information. However, this traditional approach still has its limitations in the face of complex product classification systems and detailed characterization of user preferences. Therefore, by aggregating the feature information from the neighbors of the central node in the graph and combining the aggregated information with the current representation of the central node, this paper achieves the purpose of updating the representation of the central node in the graph neural network (GNN), and enables the machine learning algorithm to learn new tasks adaptively without redesigning and training the model. The neuronal attention aggregation layer (GMFA) is proposed. Finally, the experimental comparison proves that the combination of meta-learning and graph neural network has great effectiveness and feasibility.