新闻推荐系统中用户冷启动问题的研究

Research on the Problem of User Cold-start in News Recommendation Systems

提出利用用户上下文信息,解决新闻推荐系统中用户冷启动问题的方法.通过已有用户对于新闻的点击浏览记录,提取其在不同环境中的上下文信息,并利用兴趣分类记录构建决策树分类模型.新用户到达时,提取此用户在当前环境中所带有的上下文信息并与决策树模型进行匹配,以此预测新用户的新闻浏览兴趣,并将新闻主题与用户兴趣进行匹配,进而完成新闻推荐.实验结果表明,本文提出的基于用户上下文信息的方法能够有效缓解新闻推荐系统中用户冷启动问题,用户满意度明显提高,新闻推荐结果更为人性化.

In order to solve the problem of cold start in news recommendation system,this paper proposes the way of using the newuser＇s context information. It extracts the existing user＇s context information according to his browsing history of different news,and establishes a decision tree classification model by using their interest classification record. When the newuser arrives,it will extract the context information of him under the present environment and match it with the decision tree model,which can help predict the newuser＇s browsing interest of different news,match the news topic to it,and achieve the news recommendation to him. The experimental results showthat the way of using the newuser＇s context information proposed in this paper can effectively solve the problem in news recommendation systems,improve the user＇s satisfaction,and make the results of news recommendation more humane.