基于二分网络的用户聚类电影推荐系统构建

Construction of user clustering movie recommendation system based on bipartite graph networks

针对已经存在的推荐算法中数据的稀疏性问题,提出一种基于聚类算法的二分图信任网络构造算法,通过聚类技术把项目评分相似的用户聚集起来,形成若干个用户群组,在每个群组内部通过二分图建立连接,利用信任机制在群组内部和群组间建立连接,进而构造出推荐系统。实验是在Movie Lens数据集上进行的,采用平均绝对误差(MAE)为评测指标,验证了方法的有效性,从而得出该系统使得数据稀疏性对最终推荐结果的负面影响变小。

According to the sparsity of data in the recommendation algorithm,a bipartite graph trust network based on clustering technology is proposed. This recommendation system is constructed by clustering the score similar users together,forming a plurality of user groups.In each group by bipartite graph to establish connection,through the trust mechanism between the groups and the group to establish a connection. Experiment was carried out in Movie Lens dataset, and the mean absolute error(MAE) is used as the evaluation index,the experiment verified the validity of the method,and that the system makes the conclusion that data sparsity negative effect on the final recommendation diminish.