Hybrid Recommendation Algorithm Combining User Similarity And Item Similarity
Jinping Yu, Li Yu · 2022
Aiming at the problem that the traditional recommendation algorithm cannot accurately recommend the target user in the case of sparse data, this paper proposes a collaborative filtering recommendation algorithm that improves user similarity and item similarity.Firstly,the traditional item-based collaborative filtering algorithm just takes into account the similarity between items. It also neglects the importance of auxiliary information about items. In particular, time has a great impact on the recommendation results.Aiming at this problem, this paper introduces the time function into the traditional item similarity, and then solves the problem that the number of users’ common scores in sparse data is very small and their differences in dimensions. In this paper, the traditional user similarity is improved, and the Pearson similarity and Jacarrd coefficient is weighted and combined. Finally, the recommendation results are weighted and combined to get the final recommendation results. The experimental results show that the average value of the MAE of the fusion recommendation algorithm is 2.08% lower than the traditional UserBasedCF on the Movielens dataset, 2.89 % lower than the ItemBasedCF, and 1.09 % lower than the improved algorithm proposed in literature [7].