A Sequential Recommendation Model Based on Social Behavior

Traditional social recommendation mainly combines the user’s social relationship and rating information, which can solve the problem of data sparsity in the recommendation algorithm, but ignores the fact that the user’s interests are constantly changing, and the user’s social relationship not only has a trust relationship, but also has a distrust relationship that will affect the recommendation results. Therefore, this paper proposes a sequence recommendation model based on social behavior (SBRec). First, it uses the attention mechanism to build a user dynamic interest feature model, and then according to the user’s social network, it filters out the social list that users trust and distrust, and obtains the positive and negative social impact of users through the historical behavior of friends, The user’s dynamic interest characteristics are combined with the user’s positive and negative social impact to get the recommendation results. The accuracy of the SBRec method is verified by the experiments in the actual data set Epings, and the hit rate was improved by about 6%.