摘要
为解决基于位置社交网络中地点推荐时遇到的数据稀疏、冷启动问题,提出一种改进的地点推荐方法,在协同过滤算法的基础上融合了聚类算法,考虑到用户偏好、朋友关系、位置语义等因素,在推荐时取两种算法的优点进行互补。研究的重点是相似度的计算,包括兴趣地点相似度、好友亲密度、词频-逆文档频率、余弦相似性。在Foursquare数据集上以准确率、召回率、单个主题的平均准确率作为度量依据,对提出的方法进行验证。试验证明,本方法有效提高了推荐效果。
In order to solve the data sparse and cold start in spot recommendation in the location-based social networking,an improved spot recommendation method was proposed. Based on the clustering algorithm and the collaborative filtering algorithm,the user preferences,friend relations,semantic location and other factors was taken into account.The advantages of the two methods were complemented. The focus of this research was the calculation of similarity,which included location similarity,friends intimacy measure,term frequency inverse document frequency,cosine similarity. To verify the proposed methods,precision,recall,mean average precision was used as a measure on Foursquare dataset. The results showed that the proposed method could effectively improve the recommendation effect.
出处
《山东大学学报(工学版)》
CAS
北大核心
2016年第3期44-50,共7页
Journal of Shandong University(Engineering Science)
关键词
基于位置的社交网络
协同过滤
聚类
位置推荐
location-based social network
collaborative filtering
clustering
spot recommendation