基于位置聚类和张量分解的Web服务推荐研究与应用

发布时间：2018-05-11 19:39

  本文选题：位置近邻 + 聚类　； 参考：《重庆大学》2016年硕士论文

【摘要】：随着SOA架构和Web服务相关标准的日趋成熟,全球越来越多的开发者、组织和企业成为Web服务提供商,在各Web服务平台上开发和提供功能各异的Web服务。这使得各平台上Web服务数量急剧增加,目前网络上海量的Web服务中,相似甚至相同功能的Web服务很多,如何在众多功能相似的Web服务中发现最能满足用户需求的服务并推荐给相应用户成为一个难题。基于服务质量(Quality of Service,QoS)的Web服务推荐技术可根据服务的非功能属性为用户推荐最合适的Web服务,已成为近年来服务计算领域的研究热点。其中,准确预测缺失的QoS属性值是一个难点,目前的QoS属性值预测算法大多只根据用户的服务调用历史,采用协同过滤算法进行预测,还存在预测准确率不高的问题。为解决该问题,本文对基于服务质量的Web服务推荐系统展开研究,将位置属性和访问时间上下文结合至Qo S属性值预测之中,利用张量分解模型提高了QoS属性值的预测准确度,从而获得更合理、有效的Web服务推荐结果。本文的主要内容如下:(1)分析了Web服务推荐系统的研究背景和现状,提出了本课题的主要研究内容和创新点,并对与本文主要研究内容相关的概念和主要技术进行深入的研究和分析,包括Web服务相关技术、协同过滤算法和张量分解模型。(2)提出了两种基于位置和张量分解的Web服务Qo S预测算法:TATD算法和Clust TD算法。TATD算法将用户的地理位置属性以位置近邻正则项的形式融入至张量分解模型之中,预测活跃用户在不同时间段访问各Web服务时的Qo S属性值;Clust TD算法首先根据用户和服务的位置经纬度值将用户和服务聚类成多个局部组,再分别对各局部组和全局的用户、服务和时间上下文进行张量建模和分解,最后将局部和全局张量分解的QoS预测结果进行加权组合,考虑了用户和服务的相对位置以及访问时间上下文,能获得更准确的Web服务Qo S预测值。(3)在真实的Web服务访问数据集上验证了本文提出的TATD算法和Clust TD算法的Qo S值预测性能,并通过Web服务个性化推荐原型系统的构建对本文提出的Web服务推荐新算法进行实践,验证了新算法的可行性和有效性。
[Abstract]:With the maturity of SOA architecture and Web service related standards, more and more developers, organizations and enterprises have become Web service providers all over the world, developing and providing Web services with different functions on various Web service platforms. This makes the number of Web services on various platforms increase dramatically. At present, among the Web services in Shanghai, there are many Web services with similar or even the same functions. How to find the most suitable Web services to meet the needs of users and recommend them to the corresponding users has become a difficult problem. The Web service recommendation technology based on quality of Service (QoS) can recommend the most suitable Web service according to the non-functional attribute of the service, which has become the research hotspot in the field of service computing in recent years. It is difficult to accurately predict the missing QoS attribute value. Most of the current QoS attribute prediction algorithms are only based on the user's history of service call and use collaborative filtering algorithm to predict the missing QoS attribute value. There is still a problem of low prediction accuracy. In order to solve this problem, the Web service recommendation system based on QoS is studied in this paper. The location attribute and access time context are combined into the prediction of QoS attribute value, and the prediction accuracy of QoS attribute value is improved by using Zhang Liang decomposition model. In order to obtain more reasonable and effective Web service recommendation results. The main contents of this paper are as follows: (1) the research background and current situation of Web service recommendation system are analyzed, and the main research contents and innovation points of this subject are put forward. Furthermore, the concepts and technologies related to the main contents of this paper are deeply studied and analyzed, including the related technologies of Web services. Cooperative filtering algorithm and Zhang Liang decomposition model. 2) two Web service QoS prediction algorithms based on location and Zhang Liang decomposition are proposed, namely: TATD algorithm and Clust TD algorithm. TATD algorithm takes the geographical location attribute of user as the regular term of location nearest neighbor. Into the Zhang Liang decomposition model, In order to predict the QoS attribute value of active users visiting each Web service in different time periods, first of all, the users and services are clustered into several local groups according to the location, longitude and latitude of the user and the service, and then the local groups and the global users are treated respectively. The service and time context are modeled and decomposed by Zhang Liang. Finally, the QoS prediction results of local and global Zhang Liang decomposition are combined weighted, taking into account the relative position of the user and the service and the access time context. Can obtain more accurate Web service QoS prediction value. 3) the proposed TATD algorithm and Clust TD algorithm are verified on the real Web service access data set. The feasibility and effectiveness of the new Web services recommendation algorithm are verified by the construction of the Web services personalized recommendation prototype system.
【学位授予单位】：重庆大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TP393.09;TP391.3

【参考文献】

中国期刊全文数据库 前5条

1 李贵;王爽;李征宇;韩子扬;孙平;孙焕良;;基于张量分解的个性化标签推荐算法[J];计算机科学;2015年02期

2 邹本友;李翠平;谭力文;陈红;王绍卿;;基于用户信任和张量分解的社会网络推荐[J];软件学报;2014年12期

3 邓水光;黄龙涛;吴斌;尹健伟;李革新;;一种QoS最优的语义Web服务自动组合方法[J];计算机学报;2013年05期

4 王立才;孟祥武;张玉洁;;上下文感知推荐系统[J];软件学报;2012年01期

5 刘建国;周涛;汪秉宏;;个性化推荐系统的研究进展[J];自然科学进展;2009年01期

，

本文编号：1875340