【摘要】 为解决协同过滤方法中面向新用户的冷启动问题,提出基于用户特征的混合Top-N推荐方法.通过对用户特征的聚类分析,获得特征相似用户集,根据目标新用户与不同用户集的距离、用户集对不同项目的评分来预测新用户对项目的评分,并结合传统协同过滤方法进行推荐,最后采用命中率来评价推荐的准确度.实验结果表明,该方法针对新用户有较好的推荐精度.更多还原

【Abstract】 Aiming at the cold start user problem of traditional recommendation system,this paper proposed a hybrid Top-N recommendation method which combined user feature and collaborative filtering.The method obtained the similar user set through the cluster analysis about the user feature,and predicted the ratings on items from new user basing on which is not only distance between the new user and the different user set,but also the ratings of the different project from user set.Finally,the traditional collaborative filtering is employed and the accuracy is evaluated by Hit Rate.Empirical results show our Hybrid Top-N recommender system have better accuracy for new user.更多还原