【作者】 王冰；

【导师】 胡强；

【作者基本信息】 青岛科技大学 ， 计算机技术（专业学位）， 2022， 硕士

【摘要】 随着面向服务架构(service oriented architecture,SOA)在软件开发过程中的应用和普及,互联网上涌现出大量Web服务。如何管理和组织这些服务并从众多的相似服务中查找满足用户需求的服务成为服务计算领域一个亟待解决的问题。为了提高服务查找的精确度和效率,通常采用服务聚类的方式缩减服务的查找空间。目前的服务聚类方法主要关注服务功能相似度的计算,在计算功能相似度时通常是通过主题模型对服务描述文本进行功能特征提取,然而由于服务描述文本中的单词稀疏,因此生成的服务功能向量质量不高,从而影响了聚类效果。此外,当前服务聚类研究缺乏对服务协作关系的考量,未能很好地将服务之间的协作相似度融入到聚类过程中。针对上述问题,本文提出一种融合功能语义与流程协作关系度量的服务聚类方法,主要工作和创新点如下:(1)在计算服务功能语义相似度时,提出了一种特征词提取方法WTPF。针对现有传统提词方法面对高稀疏度数据时提取特征词有效性低的问题,创新提词方法,引入词性、语义权重等属性共同表征词语的重要程度,以获取较高质量的特征词。在此基础上,通过GSDMM主题模型对提取的特征词生成功能语义向量,从而生成高质量的服务功能向量,提高了服务功能语义相似度的计算精度。(2)依据服务之间的流程协作关系建立服务流程协作网络,基于Node2Vec为流程协作网络中的服务生成流程协作向量。通过协作向量计算服务之间的协作相似度,并将服务协作相似度引入到服务聚类中,从而进一步提高服务聚类的质量。(3)构建了融合功能语义与流程协作相似度的FPK-means++聚类算法。通过调整功能语义相似度与流程协作相似度在服务相似度计算过程中的比例,获得更优的聚类效果。实验验证了本文方法的有效性和先进性。更多还原

【Abstract】 With the application and popularization of service-oriented architecture(SOA)in the process of software development,a large number of Web services have emerged on the Internet.It is still an urgent question of how to organize and manage these services and find services that meet the needs of users from similar services in the field of service computing.In order to improve the accuracy and efficiency of service search,service clustering is usually used to reduce the search space of services.The current service clustering methods mainly focus on the calculation of service function similarity.When calculating the function similarity,the topic model is usually used to extract the functional features of the service description text.However,due to the sparse words in the service description text,the quality of the generated service functional vectors is not high,which affects the clustering effect.In addition,the current research on service clustering lacks consideration of the service cooperation relationship,and fails to well integrate the cooperation similarity between services into the clustering process.To address the above problems,this paper proposes a service clustering method that integrates functional semantics and process collaboration relationship measurement.The main work and innovations are as follows:(1)The existing traditional extraction methods present low effectiveness in extracting feature words when faced with high sparse data.Aiming at the problem,this thesis innovates a word extraction method that introduces attributes include part of speech and semantic weight to jointly represent the importance of words,so as to obtain high-quality feature words.Based on this,the GSDMM topic model is used to generate functional semantic vectors for the extracted feature words,so as to generate high-quality service function vectors and improve the calculation accuracy of the semantic similarity of service functions.(2)The service process collaboration graph is established according to the process collaboration relationship between services,and the process collaboration vector is generated for the services in the process collaboration graph based on Node2 Vec.The cooperation similarity between services is calculated by the cooperation vector,and the service cooperation similarity is introduced into the service clustering,further improve the quality of the service clustering.(3)A FPK-means++ clustering algorithm that integrates functional semantics and process collaboration similarity is constructed.By adjusting the ratio of functional semantic similarity and process collaboration similarity in the process of similarity calculation to obtain a superior clustering effect.Experiments verify the effectiveness and advancement of the method in this paper.更多还原