基于用户签到行为的群组兴趣点推荐模型

Group POI Recommendation Model Based on the User Check-in Behavior

推荐系统将符合用户兴趣分布的项目推荐给用户.目前对推荐系统的研究大多集中于对个体用户进行推荐.然而在实际生活中,很多活动是由多个用户共同参与的.因此,组推荐系统逐渐成为研究的热点.本文提出一种基于用户签到行为的群组兴趣点推荐模型(Group POI recommendation model based on the User Check-in behavior,GPUC),该模型首先采用协同过滤的推荐算法挖掘组内成员可能感兴趣的项目,并基于TF-IDF(term frequency-inverse document frequency)的思想预测用户对项目的评分,生成个人推荐列表.在融合组内成员兴趣偏好时,提出一种加权混合融合策略,兼顾考虑不同组员在群组中的权重以及群组偏好差异度的大小.本文采用Gowalla网站的真实数据集验证了推荐模型的准确性,与算法HAaB相比,基于用户签到行为的群组兴趣点推荐模型GPUC的准确率提高了4. 03%,为群组用户提供了更有效的推荐.

The recommender system recommends items which are in line with user＇s interests to target users. At present, most researches on recommendation systems focus on recommending for an individual user. However, in real life, many activities are jointly participated by multiple users. The group recommendation system has gradually become a research hotspot. This paper proposes a Group POI recommendation model based on the User Check-in Behavior （ GPUC ）. First, it uses the collaborative filtering recommendation algorithm to mine the items that may be of interest to the group members, and predictes the user＇s score of a project based on the idea of TF-IDF （ predictive frequency-inverse document frequency） to generate a personal recommendation list. When aggregating mem- bers＇ preferences into group preference, this paper proposes a weighted hybrid aggregation strategy which takes into account the weight of different group members and the differences of the group members＇ preference. At the end of the paper we use the real data set of Gowalla website to verify the accuracy of the recommended model. Compared with algorithm HAaB, the accuracy of GPUC is improved by 4.03% ,providing more accurate result for group users.