基于用户聚类的二分图网络协同推荐算法

针对协同过滤推荐系统应用中存在的数据稀疏、可扩展性受限等问题,提出了一种基于用户聚类的二分图网络协同推荐算法.该算法在用户聚类阶段对二分图网络进行用户中心聚类,并获取用户聚类中心及其所在的群组,基于用户群组的评价信息为目标用户提供更广泛的推荐数据;在协同推荐阶段,围绕聚类中心及其所在群组为未评分项目完成预测评分,为用户推荐综合评分最高的Top-n项目.结果表明,该算法能够提升目标用户推荐的准确度,并能改善协同推荐的多样性.

Collaborative recommendation algorithm for bipartite networks based on user clustering

Aiming at the problems about data sparsity and limited scalability in the application of collaborative filtering recommendation system, a collaborative recommendation algorithm for bipartite networks based on user clustering was proposed. The user center clustering was carried out for the bipartite networks in the user clustering stage, and the user clustering centers and the corresponding groups were obtained. In addition, more recommendation data were provided for the target users based on the evaluation information of user group. In the collaborative recommendation stage, the prediction scoring was finished for the projects without scoring around the clustering centers and their groups, and the Top-n projects with the highest comprehensive scores were recommended for the users. The results show that the proposed algorithm can enhance the recommendation accuracy of target users, and improve the diversity of collaborative recommendation.