基于混合推荐和隐马尔科夫模型的服务推荐方法  

Recommending services via hybrid recommendation algorithms and hidden Markov model in cloud

作　　者：马建威[1,2] 陈洪辉[1] STEPHAN Reiff-Marganiec 

机构地区：[1]国防科学技术大学信息系统工程重点实验室,湖南长沙410073 [2]第三军医大学卫勤教研室,重庆400038 [3]莱斯特大学计算机科学与技术系

出　　处：《中南大学学报（自然科学版）》2016年第1期82-90,共9页Journal of Central South University:Science and Technology

基　　金：国家自然科学基金资助项目(71071160);全军后勤科研重点计划项目(BWS14J032);湖南省优秀研究生创新项目(CX2011B024);国防科技大学优秀研究生创新项目(B110502);第三军医大学人文社科基金资助项目(2015XRW10)~~

摘　　要：针对现阶段越来越多的服务开始部署于云环境,服务数量呈几何级增长,必须获取并推荐最优服务,而传统的基于内容的过滤或协同过滤方法缺乏对新用户和冗余服务的有效处理方法,提出一种在云环境下对最优服务进行有效推荐的方法。首先,分析2种协同过滤方法的优缺点,并提出改进的混合推荐算法;其次,针对常常被忽略的新用户学习策略,提出新用户偏好的确定方法;针对服务的动态变化情况,基于隐马尔科夫模型(hidden Markov model)提出一种冗余服务消解策略。最后,基于真实数据集和通过公开API获取的公共服务集进行实验。研究结果表明:所提出的算法与其他方法相比具有更高的准确度和更好的服务质量,能更有效地提高系统性能。With the increase of the number of users using web services for online activities through thousands of services, proper services must be obtained; however, the existing methods such as content-based approaches or collaborative filtering do not consider new users and redundant services. An effective approach was proposed to recommend the most appropriate services in a cloud computing environment. Firstly, a hybrid collaborative filtering method was proposed to recommend services. The method greatly enhances the prediction of the current Qo S value which may differ from that of the service publication phase. Secondly, a strategy was proposed to obtain the preferences of the new users who are neglected in other research. Thirdly, a HMM（hidden Markov model）-based approach was proposed to identify redundant services in a dynamic situation. Finally, several experiments were set up based on real data set and publicly published web services data set. The results show that the proposed algorithm has better performance than other methods.

关 键 词：协同过滤 服务选择 新用户学习 隐马尔科夫模型 冗余检测 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]