基于均衡接近度增强时间的兴趣点推荐模型  

Enhanced Time Point-of-interest Recommendation Model Based on Balanced Adjacent Degree

作　　者：陈江美 张岐山 张文德[2] 何珑[3] CHEN Jiang-mei;ZHANG Qi-shan;ZHANG Wen-de;HE Long(School of Economy and Management,Fuzhou University,Fuzhou 350108,China;Institute of Information Management,Fuzhou University,Fuzhou 350108,China;Network and Information Construction Office,Fuzhou University,Fuzhou 350108,China)

机构地区：[1]福州大学经济与管理学院,福州350108 [2]福州大学信息管理研究所,福州350108 [3]福州大学网络与信息化建设办公室,福州350108

出　　处：《小型微型计算机系统》2020年第10期2075-2081,共7页Journal of Chinese Computer Systems

基　　金：国家自然科学基金项目(61300104)资助;福建省自然科学基金项目(2018J01791)资助;赛尔网络下一代互联网创新项目(NG20170522,NGII20180509)资助。

摘　　要：针对现有的兴趣点推荐算法存在用户签到数据稀疏的问题和用户兴趣的动态变化问题,提出了基于均衡接近度增强时间的兴趣点推荐模型.首先,采用均衡接近度方法计算时间相似度,获得时间影响模型.然后,结合融入兴趣点流行度的空间影响模型,建立矩阵填充模型,将得分最高的前s个兴趣点填充进矩阵.最后,将时间因素融入到矩阵分解模型中,进行优化求解.实验结果表明,该模型更有效地缓解数据稀疏性和用户兴趣动态变化的问题,推荐性能明显优于其他的基准模型.Since the current w orks of point-of-interest recommendation exsit sparsity of check-in data and dynamic change of user interests,this paper proposed a enhanced time point-of-interest recommendation model based on balanced adjacent degree.Firstly,the method of balanced adjacent degree calculated the time similarity,obtained time influence model.Then,combined w ith the spatial influence model w hich integrates the popularity of point-of-interest,then established a matrix filling model,and filled the top-s point-ofinterest w ith the highest score into the matrix.Finally,time factor is integrated into the matrix decomposition model to optimize the solution.Experimental results show the model can effectively alleviate the problem of data sparsity and dynamic changes of user interests,and the recommendation performance outperforms other baseline models.

关 键 词：兴趣点推荐 均衡接近度 时间影响 空间影响 矩阵填充 矩阵分解 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]