基于核心用户和协同过滤的多样性推荐方法

Diversity Recommendation Method Based on Core Users and Collaborative Filtering

基于协同过滤算法的推荐系统对某一目标用户产生推荐结果时,存在追求高准确度而牺牲多样性的问题。为了改善并解决这一问题,提出在协同过滤的基础上融入核心用户的多样性推荐算法。首先,在推荐系统的所有用户数据中根据用户评价的项目数量以及评价的准确度筛选出核心用户;然后,依次计算待推荐项目的特定用户与各个核心用户两两之间的相似性,目标用户的k最近邻即根据相似性的大小排序排在前k个位置的用户;最后,将从基于核心用户得到的k个最近邻与基于用户的协同过滤推荐技术得到的前k个最近邻进行权重调整结合得到最终的推荐列表。实验结果显示,本方法在保证一定程度推荐准确性的情况下,可以有效增大系统的推荐多样性以及用户的满意度。

When a recommendation system based on collaborative filtering algorithm produces a recommendation result for a certain target user,it has the problem of seeking high accuracy at the expense of diversity.In order to solve this problem,this paper proposes a diversity recommendation algorithm based on collaborative filtering and incorporating core users.Firstly,the core users are screened out from all user data of the recommendation system according to the number of items evaluated by users and the accuracy of evaluation.Then,the similarity between the specific users and core users of the recommended project is calculated in sequence.The K-nearest neighbor of the target user is the user in the top K positions ranked according to the size of the similarity.Finally,the final recommendation list is obtained by adjusting the weights of the K nearest neighbors obtained based on the core users and the top K nearest neighbors obtained by the user-based collaborative filtering recommendation technology.Experimental results show that this method can effectively increase the diversity of recommendations and user satisfaction of the system while ensuring a certain degree of recommendation accuracy.