基于K-means和矩阵分解的位置推荐算法  

Location recommendation algorithm based on K-means and matrix factorization

作　　者：李宾[1] 周旭 梅芳[3] 潘帅宁 LI Bin;ZHOU Xu;MEI Fang;PAN Shuai-ning(College of Mathematics,Jilin University,Changchun 130012,China;Center for Computer Fundamental Education,Jilin University,Changchun 130012,China;College of Computer Science and Technology,Jilin University,Changchun 130012,China;College of Computer Science and Technology,Huazhong University of Science and Technology,Wuhan 430074,China)

机构地区：[1]吉林大学数学学院,长春130012 [2]吉林大学公共计算机教学与研究中心,长春130012 [3]吉林大学计算机科学与技术学院,长春130012 [4]华中科技大学计算机科学与技术学院,武汉430074

出　　处：《吉林大学学报（工学版）》2019年第5期1653-1660,共8页Journal of Jilin University:Engineering and Technology Edition

基　　金：国家自然科学基金项目(61806083,61872158);吉林省科技厅优秀青年人才基金项目(20190103051JH);吉林省教育厅“十三五”科学技术项目(JJKH20190161KJ)

摘　　要：采用矩阵分解方法为模型量化用户对未知位置的签到次数,利用改进的融合用户间签到和好友关系的相似度计算方法计算用户间的相似度,通过聚类方法将用户进行群组划分,最终提出结合聚类和矩阵分解的方法实现个性化位置推荐。相比于基于用户协同过滤和基于矩阵分解推荐算法,本文算法在位置推荐召回率和准确率上均有提高,同时,在推荐运行时间上也优于其他算法。In this paper, a novel POI recommendation algorithm based on clustering and matric factorization is proposed. The matrix factorization method is used to quantify the user’s number of check-in for unknown location. The clustering method is used to divide the users into groups. In addition,the improved similarity calculation method is used to calculate the similarity among users. The experiment results on Yelp dataset show that the proposed method can improve the location recommendation recall rate and accuracy rate,demonstrating the better performance of our method compared to other methods.

关 键 词：计算机应用 矩阵分解 K均值聚类 位置推荐 基于位置社交网络 

分 类 号：TP391.3[自动化与计算机技术—计算机应用技术]