基于服务关联的服务推荐和发现方法研究

发布时间：2018-09-19 14:50

【摘要】：面向服务的计算(Service Oriented Computing,简称SOC)是当前软件领域备受关注的热门主题之一。SOC倡导以标准的方式支持系统的开放性,它所提供的服务协同和管理改善了软件产品复杂的业务系统,提高了软件系统的生产效率。面向服务的架构(Service Oriented Architecture, SOA)使得分布式应用具有更好的灵活性和复用能力。SOA的主流实现方式是Web服务技术,随着面向服务架构技术的大量应用,当前Web服务的数量正在以超线性的速度增长。面对快速膨胀的Web服务资源,用户如何才能方便、准确的从大量的服务资源中找到自己需要的服务成为了当前工业界和学术界的一大挑战。 Web服务推荐和发现技术的出现为服务发现和查找难问题的解决提供了一个方向。现有的热门推荐技术在电子商务行业的商品推荐中应用较为广泛,但是由于Web服务的异构性和用户需求的多样性,传统的推荐技术简单地应用到Web服务推荐中往往推荐的准确度较低。因此,如何把传统推荐技术运用在Web服务推荐中,并提高推荐的准确度是当前相关研究的难点之一。现有技术大多从Web服务本身所蕴含的描述信息及其用户之间的相似度出发为用户推荐服务。这种方法大多忽略了服务之间的内在关联和兼容性,推荐过程没有结合用户自己的Web服务进行考虑,使得推荐的Web服务不能保证与用户自己的Web服务组合使用,造成服务资源的浪费,未能很好地满足SOC软件重用的初衷。服务的推荐技术可以在用户的功能需求还不明确的情况下完成推荐,而服务发现方法则为用户找到特定功能的服务提供了便捷。目前由于Web服务的描述文档缺乏语义信息,服务的发现在准确度和完备性上一直存在不足。基于本体论的Web服务发现研究还不成熟,而现有基于规则、聚类方法和文本向量空间模型的研究取得了较好的效果。基于Web服务之间存在着相似度上的关联,本文从Web服务聚类问题和文本向量空间模型的构建入手,对目前Web服务聚类效果差的问题进行研究,指出现有聚类方法和Web服务相似度计算方面的不足,并针对性地提出新的解决方法。本文的主要研究内容和创新点如下： (1)针对服务推荐存在的问题,提出了基于关联规则挖掘的Web服务推荐方法(Services Recommending Method Based on Association Rule Mining,简称RecARM)。RecARM利用Web服务组合的历史记录构建Web服务间的关联规则,挖掘服务之间潜在的关联关系,利用用户自有的Web服务为用户进行推荐,帮助完善和优化用户的服务组合。实验结果表明,改进后的服务推荐结果相较于常规的推荐方法在稳定性和准确度上均有提高。该方法有效利用了Web服务所特有的历史组合记录的数据,为推荐提供了依据。 (2)针对服务的发现引入新的聚类算法,根据Web服务间的相似性关联提出了一种基于聚类的Web服务发现方法。本文引入改进的ISODATA聚类算法,该方法有效解决了Web服务发现过程中聚类数量无法确定的问题,并降低了异常数据对推荐结果产生的干扰和影响。文本所提出的方法为服务的推荐和发现提供了新的方法和思路。
[Abstract]:Service Oriented Computing (SOC) is one of the hot topics in the software field. SOC advocates the openness of support systems in a standard way. The service coordination and management provided by SOC improves the complex business systems of software products and the productivity of software systems. Service Oriented Architecture (SOA) makes distributed applications more flexible and reusable. The mainstream implementation of SOA is Web services technology. With the extensive application of Service Oriented Architecture (SOA), the number of Web services is increasing at a superlinear rate. In order to conveniently and accurately find their own needs from a large number of service resources has become a major challenge in the current industry and academia.
The emergence of Web service recommendation and discovery technology provides a direction for solving the problem of service discovery and discovery.
Current popular recommendation technologies are widely used in the commodity recommendation of e-commerce industry. However, due to the heterogeneity of Web services and the diversity of users'needs, traditional recommendation technologies are often used to recommend Web services with low accuracy. Improving the accuracy of recommendation is one of the difficulties in current research. Most of the existing technologies are based on the description information contained in Web services and the similarity between users. Row considerations make the recommended Web services not guaranteed to be used in combination with the user's own Web services, resulting in a waste of service resources and failing to meet the original intention of SOC software reuse.
Service recommendation technology can accomplish recommendation when the user's functional requirements are not clear, while service discovery method provides convenience for users to find services with specific functions. The research of EB service discovery is not mature, but the existing research based on rules, clustering methods and text vector space model has achieved good results. Based on the similarity between web services, this paper starts with the clustering problem of Web services and the construction of text vector space model, and advances the problem of poor clustering effect of web services. The main contents and innovations of this paper are as follows:1.
(1) Aiming at the problems of service recommendation, a Web service recommendation method based on association rule mining (RecARM) is proposed. RecARM uses the history of Web service composition to construct association rules between Web services, mining the potential association between services, and utilizing users. The experimental results show that the improved service recommendation results are more stable and accurate than the conventional recommendation methods. This method effectively utilizes the data recorded by the unique historical composition of Web services and provides a basis for recommendation.
(2) Introduce a new clustering algorithm for service discovery, and propose a clustering-based Web service discovery method according to the similarity relationship between Web services. This paper introduces an improved ISODATA clustering algorithm, which effectively solves the problem that the number of clusters can not be determined in the process of Web service discovery, and reduces the recommendation result of abnormal data. Interference and impact.
The method proposed in this paper provides new methods and ideas for service recommendation and discovery.
【学位授予单位】：浙江工商大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TP393.09

【参考文献】