【作者】 李威；

【导师】 孙其博；

【作者基本信息】 北京邮电大学 ， 计算机科学与技术， 2015， 硕士

【摘要】 信誉度用来描述服务提供方发布的QoS和真实值之间的可信度。其度量标准是依据用户使用服务后,根据获取到的真实QoS与公告值的差别,以及用户主观感受所给出的反馈等级综合得到。准确的信誉度对于用户进行服务选择,有非常重要的作用。尽管已有的很多方法能够有效地提高服务信誉度度量的准确性,但其普遍忽略了用户上下文对信誉度度量的影响。为此,本文针对用户上下文对反馈等级的影响以及如何弱化用户上下文影响进行了深入的研究,通过进行用户上下文聚类以及弱化用户上下文对反馈等级的影响,来对反馈等级进行预处理。本文面向不同的应用场景,提出了两种基于用户上下文的Web服务信誉度度量方法：1)针对用户上下文对用户反馈等级产生不同影响的问题,提出了一种基于用户上下文聚类的Web服务信誉度度量方法。该方法首先对用户上下文建模,并对其进行形式化定义及正规化处理；然后,对用户上下文进行聚类,旨在用一个综合量化值来表征用户上下文级别；最后,使用同一量化值的反馈等级来度量信誉度,进而获得不同用户上下文中的信誉度度量结果。实验结果表明,在用户反馈等级受用户上下文影响较大的场景下,该方法能够有效提高信誉度度量的准确性。2)针对已有方法在信誉度度量过程中仅采用单一用户上下文中的反馈等级而造成反馈等级使用率低的问题,提出了一种基于用户上下文敏感度的信誉度度量方法。该方法首先提出用户上下文敏感度概念,并对用户上下文敏感度进行度量；然后,根据用户上下文敏感度对反馈等级中的用户上下文影响进行弱化,将不同用户上下文中的反馈等级转化为统一参考级别；最后,根据弱化后的反馈等级度量用户相似度,并以此获得服务信誉度的度量结果。实验结果表明,在用户对服务频繁访问的场景下,该方法能够有效提高信誉度度量的准确性。更多还原

【Abstract】 Reputation is used to describe the credibility and reliability of QoS claimed by service provider. Its measurement is based on the feedback ratings that users give to the service after invoking. Feedback ratings reflect users’subjective feeling and objective effect according to the acquired QoS. Accurate reputation value is very important for users to choose services. Therefore, Web service reputation measurement has become a hot research topic in the field of service computing.Although many existing approaches can effectively improve the accuracy of Web service reputation measurement, they generally ignore the impact of user’s context on reputation measurement. For this purpose, we conduct deep research on the impact of user’s context and how to weaken the impact. We preprocess the feedback ratings by clustering user’s context and weakening its impact on feedback ratings. For different application scenarios, this paper proposes two approaches of Web service reputation measurement based on user’s context:1) Aiming at the problem that different user’s context may affect user’s feedback ratings differently, this paper proposes Web service reputation measurement approach based on user’s context clustering. This approach firstly models, formalizes and normalizes user’s context. Then, it clusters user’s context in order to use a comprehensive quantitative value to represent the level of user’context. Finally, this approach uses feedback ratings with the same context to measure Web service reputation, so that we can get different reputation value with different user’s context level. Experiment results show that, in such scenery that user’s context largely affects user’s feedback rating, our approach can effectively improve the accuracy of reputation measurement.2) Aiming at the problem that the existing approaches only use feedback ratings with the same context, which leads to low utilization of feedback ratings, this paper proposes Web service reputation measurement approach based on user’s context sensitivity. This approach firstly proposes the concept of user’s context sensitivity and measure it for users. Then, according to the acquired user’s context sensitivity, it weakens the user’s context impact on users’feedback ratings, and transfers feedback ratings into a unified reference context level. Finally, this approach measures users’ similarities according to the weakened feedback ratings, and then gets the service reputation value. Experiment results show that, in such scenery that users frequently invoke some fixed services, our approach can effectively improve the accuracy of reputation measurement.更多还原