A Collaborative Filtering Methods Based on Expected Utility

In recent years, there has been a variety of improved collaborative filtering methods, which promote accuracy of recommendation by analysing too much user or item information like individual characteristics, application scenarios and so on. This increases the time complexity of the algorithm and leads to low efficiency of recommendation. Therefore, how to improve the quality of recommendation through analysing a small amount of effective information has attracted more attention in the research of collaborative filtering methods. This paper proposes a collaborative filtering method based on expected utility from the perspective of the economic significance of user rating. Firstly, the method establishes the corresponding similarity criterion and calculates the similarity between users or items by scoring expectation. Then, the nearest neighbour set is calculated and the final prediction is got. Finally, our method is validated by the experimental results using movielens-1m data set. The results show that the collaborative filtering method proposed in this paper is simpler and more feasible compared with the traditional IBCF, KNN and other mainstream methods, and can improve the efficiency and quality effectively of recommendation only using less information. This work has significance both in theoretical research and practical application value.