基于位置的移动推荐系统效用评价研究  

A Survey of Evaluation for Location-Based Mobile Recommender Systems

作　　者：孟祥武[1,2] 梁弼 杜雨露[1,2] 张玉洁[1,2] MENG Xiang-Wu;LIANG Bi;DU Yu-Lu;ZHANG Yu-Jie(Beijing Key Laboratory of Intelligent Telecommunications Software and Multimedia(Beijing University of Posts and Telecommunications),Beijing 100876;School of Computer Science,Beijing University of Posts and Telecommunications,Beijing 100876)

机构地区：[1]智能通信软件与多媒体北京市重点实验室(北京邮电大学),北京100876 [2]北京邮电大学计算机学院,北京100876

出　　处：《计算机学报》2019年第12期2695-2721,共27页Chinese Journal of Computers

摘　　要：近年来,基于位置的移动推荐系统已经成为个性化推荐服务研究领域的热门课题之一.如何在有限数据集和用户量的情况下,采用恰当的评价方法和指标来有效评估推荐系统性能,已成为移动推荐系统研究的关键任务.本文首先概括分析了基于位置的移动推荐系统效用评价在国内外的研究进展,并与传统推荐系统效用评价进行比较;然后重点从数据集、评价方法、评价指标三方面来对基于位置的移动推荐系统进行详细分析、比较和总结,并发现一些特殊的评价方法和评价指标;同时提出一种基于位置的移动推荐系统四层评价体系,它合理地将模型、数据集、评价方法、评价指标等有机地结合起来,并恰当呈现出这些评价要素之间的相互关系;最后对基于位置的移动推荐系统效用评价的有待深入研究的问题及发展趋势进行展望,并得出一些相关结论.With the rapid development of technologies such as mobile terminals,wireless communications and Internet of Things,location-based mobile recommender systems have been widely used and become increasingly popular in recent years,and this kind of recommender system has become a hot issue in the research area of personalized recommendation services.Its goal is to recommend appropriate items to users at any time and any place,so that users can enjoy satisfactory mobile services anywhere,anytime.Moreover,the recommended performance determines the service quality of the system,so location-based mobile recommender systems have a key task that is how to use appropriate evaluation methods and metrics to effectively evaluate their performance in the case of limited datasets and users.In this paper,we review the progress of research on the evaluation of location-based mobile recommender systems in recent years.Specifically speaking,we first summarize the current research status of evaluation for location-based mobile recommender systems both at home and abroad,and compare them with traditional recommender systems from three aspects of evaluation methods,evaluation metrics and datasets.Secondly,we provide an overview of datasets,not only to clarify the importance of datasets in evaluating location-based mobile recommender systems,but also to point out the relationship between datasets and recommendation models,evaluation methods and evaluation metrics.At the same time,the current usage circumstances of four common datasets,namely Foursquare,Yelp,Gowalla and DoubanEvent are compared and analyzed.Thirdly,we elaborate on the three evaluation methods(i.e.the subjective evaluation method,the objective evaluation method and the subjective-objective evaluation method),which are often used in location-based mobile recommender systems from the aspects of definition,process,application,advantages and disadvantages.In addition,we draw a conclusion that the subjective-objective evaluation method can make full use of the advantages of sub

关 键 词：地理位置 移动推荐系统 数据集 评价方法 评价指标 评价体系 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]