基于马尔科夫过程和多属性决策的云服务个性化推荐

发布时间：2018-03-30 15:20

本文选题：云服务　切入点：马尔科夫　出处：《北京邮电大学》2015年硕士论文

【摘要】：个性化推荐在信息系统以及电子商务领域已经是非常成熟的技术了,而且表现出色。但是,数据量的大幅增长,数据类型的多样性,不仅给数据的存储和管理带来了困难,同样也成为了个性化推荐的发展瓶颈。然而,作为一种新型的计算模式和服务提供模式,云计算为解决海量数据的管理,提高推荐速度等提供了一种可行的方法。但是,云计算的出现则给个性化推荐带来了新的问题。根据用户的个性化需求,为用户推荐满足其个性化需求的服务是个性化推荐的根本目的。现有的针对云服务的个性化推荐或选择的方法一定程度上解决了用户查找服务难,选择合适服务难的问题,但是这些方法依然存在着一定的不足。首先,现存的很多方法,例如基于协同过滤的推荐,虽然强调用户的个性化需求,但是没有明确的指明用户的个性化到底是什么；其次,用户的需求并非一成不变,而是动态变化的,但是大多方法并没有明确清楚的区分用户的当前需求和非当前需求的不同；第三,随着云计算影响的深入,用户的需求不单纯的集中在功能的方面,与传统PC应用相比,云平台上的应用的实时性能将会成为用户关注的重点,所以用户的需求包括功能需求和性能需求两个方面。第四,用户同时有的需求不止一个,但是现有的很多方法都没有区分对待这些需求。而现有的方法中能够综合考虑上述四个问题的更少,因此,本文提出了一种基于马尔可夫模型和多属性决策的云服务个性化推荐模型,希望能够在一定程度上解决所提出的问题。通过对用户使用服务习惯的分析,将用户使用服务的过程抽象为马尔科夫模型。此外,针对服务存在的多个属性,提出了一种多效用合并的方法。通过实验,指标所显示的结果表明本文所提出的方法达到了在一定程度上解决所提出问题的目的,并且在一定程度提高了推荐性能。
[Abstract]:Personalized recommendation is a very mature technology in the field of information system and electronic commerce, and it has done well. However, the huge increase of data volume and the diversity of data types not only bring difficulties to the storage and management of data. However, as a new model of computing and service provision, cloud computing provides a feasible way to solve the management of massive data and improve the speed of recommendation. However, as a new model of computing and service delivery, cloud computing provides a feasible way to improve the speed of recommendation. The emergence of cloud computing brings new problems to personalized recommendation. According to the individual demand of users, the basic purpose of personalized recommendation is to recommend the services to meet their personalized needs. The existing methods of personalized recommendation or selection for cloud services to some extent solve the difficulty of finding services for users. It is difficult to choose the right service, but these methods still have some shortcomings. First, many existing methods, such as collaborative filtering of recommendations, although emphasizing the personalized needs of users, Second, the user's needs are not fixed, but dynamic, but most of the methods do not clearly distinguish the user's current needs and non-current needs. Third, with the impact of cloud computing, the needs of users are not simply focused on the functional aspects, compared with traditional PC applications, the real-time performance of applications on the cloud platform will become the focus of attention. So the requirements of users include two aspects: functional requirements and performance requirements. Fourth, users have more than one requirement at the same time. But many of the existing methods do not distinguish between these requirements, and fewer of the existing methods are able to take these four issues into account, so, This paper presents a personalized recommendation model for cloud services based on Markov model and multi-attribute decision making, which hopes to solve the problems to some extent. The process of users using services is abstracted as Markov model. In addition, a multi-utility merging method is proposed for the existence of multiple attributes of services. The results show that the proposed method achieves the purpose of solving the problem to a certain extent and improves the performance of recommendation to a certain extent.
【学位授予单位】：北京邮电大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：TP391.3;TP393.09

【参考文献】