内容增强与时间匹配的兴趣点推荐方法

Point-of-interest recommendation method based on content enhancement and time matching

针对目前兴趣点推荐对影响用户签到决策因素的全面分析不足,很少将评论情感识别与时间因素分析同时融入推荐框架等问题,探索一种兴趣点推荐的改进方法。在分析地理位置与社交关系对用户签到行为影响的基础上,将评论内容与签到时间两个因素同时融入到推荐中,集成兴趣点的类别与流行度信息。使用自然语言处理技术挖掘评论文本的情感倾向,用于调整用户签到偏好估计;通过建模用户活动时间规律与兴趣点时间流行度,将时间因素融入推荐模型,向用户推荐Top-N兴趣点。基于Foursquare真实签到数据集的对比实验结果表明,与目前主流方法相比,该模型在精确率和召回率上均有明显提升。

Aiming at the existing problems of current point-of-interest(POI)recommendation,such as the insufficiency of comprehensive analysis of factors affecting users’check-in decision-making,and the lack of integration of comment sentimental recognition and time factor analysis into the recommendation framework at the same time,an improved method of POI recommendation was explored.Based on the analysis of the influence of geographical location and social relationship on users’check-in behakvior,the two factors of comment content and check-in time were integrated into the recommendation at the same time,and the category and popularity information of POI were integrated.Natural language processing technology was used to mine the sentimental orientation of comment and the estimation of user check-in preference was adjusted.By modeling user activity time rule and POI time popularity,the time factor was integrated into the recommendation model to recommend the Top-N points of inte-rest to the user.The comparison experiment based on Foursquare real check-in data set shows that,compared with the current mainstream methods,the proposed model significantly improves precision and recall.