面向兴趣主题的个性化好友推荐

Topic-oriented recommendation based on user's interest

个性化的好友推荐是促进社交网络服务不断提高的重要途径,在大规模的社交网络环境中,准确地为用户推荐兴趣主题相似的好友能够使得用户的粘性更强,然而海量数据的稀疏性使得现有的大多数社交网络都不能够准确根据用户间兴趣的相似性进行好友推荐。为此,提出一种面向用户兴趣主题的个性化好友推荐方法(ITOR)。该方法首先采用k-core分析法提取用户的兴趣主题,在拥有相似兴趣主题的基础上结合用户属性信息,通过先验概率计算出有相同属性信息的用户成为好友的概率,进一步强化推荐结果的准确性和满意度。最后,通过爬取2015年9月份发布的新浪微博数据进行实验分析,验证了本推荐方法的有效性。

Personalized friends recommendation is animportant wayto promote the service quality of social networks.In alarge-scale social network,accurately recommending friends with similar themes can make usersstickier,but the sparseness of the massive data makes most of the existing social networks cannot accurately make friend recommendations based on similarities in the interests of users.To address this issue,apersonalized topic-oriented recommendation method based on user＇s interest（ITOR）is proposed.Firstly,the topics of user＇s interests are extracted by K-core analyzing method.Secondly,combining the users＇ attributes information,the probability of being good friends can be computed.The satisfaction and accuracy of recommendation resultsare further enhanced with this probability.Finally,we verify the effectiveness of the proposed method by crawling SinaWeibo data released in September 2015.