【作者】 罗喜军； 王韬丞； 杜小勇； 刘红岩； 何军；

【Author】 Luo Xijun~1,Wang Taocheng~1,Du Xiaoyong~1,Liu Hongyan~2,and He Jun~1 1(School of Information,Renmin University of China,Beijing 100872) 2(School of Economics and Management,Tsinghua University,Beijing 100084)

【机构】 中国人民大学信息学院； 清华大学经济管理学院；

【摘要】 个性化推荐系统的目标是推荐最合适的资源给最需要的用户,这种推荐多数是基于用户的一些历史行为而做出的.如果有足够的历史记录,协同过滤推荐方法往往比其他推荐方法要好.然而协同过滤方法存在严重的冷启动问题,即当有新的用户、新的资源时,协同过滤就无法完成推荐过程.针对冷启动问题,提出了一种新的方法,核心思想是先构造出用户和资源的类别模型,构造出"用户资源对"来标记出用户感兴趣的资源.而对于新的用户,根据其一些重要的属性特征,把他分到对应的用户类别,对于新的资源,利用贝叶斯分类方法,把他分到对应的资源类别模型.实验证明,该方法在一定程度上解决了冷启动问题.更多还原

【Abstract】 The goal of a recommender system is to suggest items of interest to a user based on historical behavior of a community of users.Given enough detailed history,collaborative filtering(CF) often performs as well or better than any of other recommendation methods.However,in cold-start situations, i.e.when recommendation needs to be done for a new user,or a new item is recommended to users,it is difficult for the CF method to do recommendation well.In this paper,a new method is proposed to solve this problem.The key idea of this approach is to design two suitable models that first construct the user-resource pairs to category resources,and then use Bayesian learning algorithm to put a new resource into a category.In this way,the newly classified resources can be recommended to users of suitable category.The experimental results show that the approach can solve the cold-start problem to some extent.更多还原