应用聚类分析的协同过滤推荐算法

The Collaborative Filtering Recommendation Algorithm Based on Clustering

在大多数推荐系统中,由于用户评分数据的稀疏性,使得相似度误差较大并影响推荐结果的准确性。为了解决用户评分数据的稀疏性问题,本文提出了一种基于聚类的协同过滤推荐算法。该算法依据项目的特征和用户对项目的评分,生成用户项目-特征兴趣矩阵,使用用户项目-特征兴趣矩阵对用户进行聚类,将同类用户对项目的平均评分作为未评分项目的估计评分填充用户-项目评分矩阵,再计算项目相似度并产生推荐结果。通过实验证明,在选取不同最近邻数量的条件下,提出的基于聚类的协同过滤推荐算法相比传统算法,推荐精度和准确度有一定程度的提高。

Most recommendation systems have large similarity error because of sparsity of user rating data,which affects the accuracy of recommendation results.In order to solve the sparsity problem of user rating data,this paper proposes a collaborative filtering recommendation algorithm based on clustering.The idea of the algorithm is to generate the user project feature interest matrix according to the characteristics of the project and the user’s rating of the project,and then use the user project feature interest matrix to cluster the users.Finally,the similarity of items is calculated and the recommendation result is generated.Experimental results show that,compared with the traditional algorithm,the proposed algorithm improves the recommendation accuracy and accuracy under the condition of selecting different number of nearest neighbors.