LBSN上基于兴趣圈中社会关系挖掘的推荐算法

发布时间：2018-01-19 09:41

本文关键词： 兴趣点推荐兴趣圈社会关系冷启动　出处：《计算机研究与发展》2017年02期 　论文类型：期刊论文

【摘要】：随着带有GPS定位功能的智能手机越来越普遍,人们喜欢分享他们的地理位置或者通过评论某个地方的商品从而留下用户的足迹,这引发了以共同的兴趣点(POIs)为中心,基于地理位置信息的社交网络研究(location based social network,LBSN).社交网络中的一类典型应用是推荐系统,而推荐系统中最常见的问题是冷启动,即在用户很少点评商家或分享评论时如何为他推荐感兴趣的商家.为解决冷启动问题,提出了一种在社交网络中基于兴趣圈的社会关系挖掘推荐算法.兴趣圈是由所有访问某一类别商品的用户群及他们之间的社会关系构成的社交联系,不同的用户访问同一类别商品表明他们对此类别具有相似兴趣.该方法在传统矩阵分解模型的基础上考虑不同的兴趣圈上的社会关系,使用的社会关系包括朋友关系(显性关系)和相关专家(隐性关系),并用它们作为规则化项来优化矩阵分解模型.实验数据集来自第5届Yelp挑战赛和自己爬取的Foursquare数据集,提出的方法与已有模型进行了充分的实验对比分析,结果表明,我们的模型特别是在解决冷启动问题方面优于多种现有的方法.
[Abstract]:As with GPS positioning function of intelligent mobile phone is becoming more and more common, people love to share their location or through comment somewhere goods leaving the user footprint, this led to the common point of interest (POIs) as the center, geographic location information based on the social network (location based social network, LBSN). A typical application in social network is the most common recommendation system, recommendation system is the problem of cold start, the users rarely comment or comment on how to share business as he recommended a business interest. To solve the cold start problem, put forward a kind of interest in social networks based on social circle the relationship between mining recommendation algorithm. Interest group is composed of all access to a category of users and their social relations, social contact, different users visit the same category of commodities that they This category has similar interests. Taking into account of the different interest circle of social relations based on traditional matrix decomposition model, the use of social relations include friends (explicit relation) and experts (relation), and they are used as the regularization term to optimize matrix decomposition model. The experimental data set from fifth the Yelp challenge and himself taken from the Foursquare data set, the proposed model and method are analyzed, experimental results show that the full comparison, our model especially in existing methods to solve the cold start problems better than many.

【作者单位】：中国科学技术大学计算机科学与技术学院;科大讯飞股份有限公司大数据研究院;武汉理工大学计算机科学与技术学院;温州大学瓯江学院计算机科学与技术学院;
【基金】：国家自然科学基金项目(61202171) 浙江省自然科学基金项目(LY15F020020) 国家“八六三”高技术研究发展计划基金项目(2015AA015403)~~
【分类号】：TP391.3
【正文快照】： This work was supported by the National Natural Science Foundation of China(61202171),the Natural Science Foundation ofZhejiang Province(LY15F020020),and the National High Technology Research and Development Program of China(863Program)(2015AA015403).(xi

【相似文献】