Recommendation Model Based on Social Homogeneity Factor and Social Influence Factor

In recent years, more and more recommendation algorithms incorporate social information. However, most social recommendation algorithms often only consider the social homogeneity factor between users and do not consider the social influence factor. To make the recommendation model more in line with the real-life situation, this paper proposes a novel graph attention network to model the homogeneity effect and the influence effect in the user domain. Besides, we also extended this idea to the item domain, using information from similar items to alleviate the problem of data sparsity. Also, considering that there will be interactions between the user domain and the item domain, which together affect the user’s preference for the item, we use a contextual multi-armed bandit to weigh the interaction between the two domains. We have conducted extensive comparative experiments and ablation experiments on two real public datasets. The experimental results show that the performance of our proposed model in the rating prediction task is better than other social recommendation model.