基于混合聚类与融合用户属性特征的协同过滤推荐算法

Collaborative filtering recommendation algorithm based on hybrid clustering and fusion of user attribute features

针对协同过滤推荐算法存在的推荐质量低、推荐效率低、冷启动等问题,提出一种基于混合聚类与融合用户属性特征的协同过滤推荐算法。根据用户属性信息,建立Canopy+K-means的混合聚类模型,采用该模型对所有用户进行聚类;生成多个聚类簇,在每个簇中结合用户属性特征,形成一种新的相似度计算模型,通过该模型找到目标用户的最近邻居,以此产生推荐列表进而实现推荐。在MovieLens数据集上进行的实验结果表明,此算法能够在提高推荐效率和推荐准确性的同时缩短算法运行时间,解决冷启动问题。

In allusion to the problems of low recommendation quality,low recommendation efficiency and cold startup in the collaborative filtering recommendation algorithm, a collaborative filtering recommendation algorithm based on hybrid clustering and fusion of user attribute features is proposed. A Canopy+K-means hybrid clustering model is established according to the user attribute information,which is used to cluster all users,so as to generate multiple clustering clusters. The user attribute feature is combined in each cluster to form a new similarity calculation model,by which the nearest neighbors of the target user is found,so that the recommendation list is generated to achieve the recommendations. The experimental results produced on the MovieLens datasets show that this algorithm can shorten the algorithm operation time and solve the cold start problem while improving the recommendation efficiency and recommendation accuracy.