针对目前大多推荐系统中使用的协同过滤算法都需要有显示的用户反馈的问题,提出一种在隐式反馈推荐系统中使用聚类与矩阵分解技术相结合的方法,为用户提供更好地推荐结果。其结果是由基于用户历史购买记录的隐式反馈产生的,不需任何显式反馈提供的数据。采用高维的、无参数的分裂层次聚类技术产生聚类结果,根据聚类的结果为每个用户提供高兴趣度的个性化推荐。实验结果表明,在隐式反馈的情况下该方法也能有效获得用户偏好,产生大量的高准确度推荐。%Aiming at the problem that most collaborative filtering algorithms require explicit user feedback ,a combination me‐thod of clustering and matrix decomposition in implicit feedback was proposed to provide users with better recommendation re‐sults ,and the results were generated using only implicit feedback based on users’ purchase history without requiring any parame‐ters from explicit feedback .A high dimensional ,parameter‐free ,divisive hierarchical clustering technique was used to produce clustering results and personalized recommendations were provided to users based on the clustering results .Finally ,experimental results demonstrate effective user preference can be obtained and a high percentage of recommendation with high ratings can be generated while using only implicit feedback through this method .
展开▼