融合用户兴趣的协同过滤算法

A Collaborative Filtering Algorithm Integrated with User Interests

传统的协同过滤算法存在三个问题:一是推荐初始阶段的冷启动问题;二是评分矩阵的数据稀疏问题;三是近考虑评分导致的相似度衡量不准确问题。三个问题导致用户相似性衡量准确性降低,最终导致推荐精准度的下降。本文考虑将网站或应用中挖掘到的用户兴趣,融入到协同过滤的用户相似性计算中。一方面可以解决协同过滤中的数据稀疏和冷启动的问题,另一方面也可以提高预测推荐的精准性。因此,本文提出了融合用户兴趣的协同过滤算法CFUI。CFUI改进了协同过滤算法中用户相似度的评估方法,在其中加入用户间兴趣的相似度。本文进行了融合参数μ最优取值的实验,随后通过实验证实了融合用户兴趣的算法CFUI比未融合用户兴趣的两个算法(UserCF和ItemCF)能够取得更小的推荐MAE,即CFUI的推荐效果更精确。

Traditional collaborative filtering algorithms have three problems.The first is the cold start problem in the initial recommendation stage;the second is the data sparse problem of the scoring matrix;and the third is the inaccuracy of similarity measurement caused by considering the scores only.These three problems lead to a decrease in the accuracy of user similarity measurement,which ultimately causes a decrease in recommendation accuracy.This paper considers incorporating user interest data into collaborative filtering algorithms.On one hand,it can solve the problem of data sparseness and cold start;on the other hand,it can improve the accuracy of similarity evaluation between users.Therefore,this paper proposes a Collaborative Filtering algorithm Integrated with User Interests(CFUI)that improves the evaluation method of user similarity.Experiments were carried out to find the optimal integration parameter.The simulation results show that the MAE of CFUI is smaller than that of UserCF and ItemCF,which indicates that CFUI can achieve more accurate recommendations.