Network Embedding Based Collaborative Filtering Model Equipped With User Purchase Motivation and Potential Interactions

Abstract

The user-item interactions in recommender systems can be modeled as a bipartite graph. With the bipartite graph, numerous researches have been devoted to learning the representation of the user’s fine-grained preferences to improve the performance of recommendation. However, the existing work cannot fully encode high-order collaborative interaction. To address the above problem, we propose a network embedding based collaborative filtering model equipped with user purchase motivation and potential interaction. Experimental results on three real datasets demonstrate that the proposed model outperforms the competitive baselines and significantly improves the recommending performance.