活动社交网络EBSNs上冷启动推荐方法

Cold-start recommendation method in event-based social networks

活动社交网络(event-based social networks,EBSNs)为用户提供了方便的组织、参加和分享社交活动的平台,由于冷启动用户/活动缺少丰富的历史数据,EBSNs上冷启动用户/活动的推荐是难点问题。面向EBSNs给出了包含活动、主办方和用户的表示模型,提出了向冷启动用户推荐活动,以及将冷启动活动推荐给用户的模型和基于随机游走的节点重要度计算方法。在分析用户参加活动行为模式的基础上,使用真实的EBSNs平台-豆瓣同城验证所提方法的有效性。提出的向冷启动用户推荐活动方案与群组活动推荐相比,评价指标有了明显的提升。该研究构建的向冷启动用户推荐活动,以及将冷启动活动推荐给用户的模型能有效解决活动社交网络上冷启动用户/活动的推荐问题。

Event-based social networks(EBSNs)provide convenient platforms of organizing,participating and sharing social events to users.This paper studies the recommendation of cold-start users/events in EBSNs.Due to the lack of rich historical data of cold-start users/events,the recommendation of cold-start users/events in EBSNS is a difficult problem.This work proposed EBSNs representation model containing event,sponsor and user.Furthermore,the models were proposed for recommending events to cold-start users or users with cold-start events,and the method of calculating node importance based on random walk was proposed.The effectiveness of this method was verified with the real EBSN platform,Douban Event.Compared with group event recommendation,the evaluation index of the proposed event recommendation scheme for cold-start users has been significantly improved.The model constructed to recommend events to cold-start users and to recommend cold-start events to users can effectively solve the recommendation problem of cold-start users/events in EBSNs.