聚类分析及关联挖掘在Web服务组合中的应用研究

Clustering analysis and association mining for Web service composition

从有助于实现 Web 服务的自动发现及最优选择的角度,从聚类分析和关联挖掘两个方面研究了数据挖掘技术在 Web 服务组合领域中的应用。仿真研究表明:根据服务之间的语义相似性,包括基本描述相似性、功能属性相似性以及社会关系相似性,对服务注册库中的原子服务进行聚类分析,在此基础上进行服务的自动查找,可在很大程度上降低服务的查找空间,提高服务的匹配效率;对服务执行日志中的历史记录进行分析,挖掘服务组合的模式以及服务之间的关联关系,结合服务质量进行服务的最优选择,可在一定程度上提高合成服务的执行成功率。

The application of data mining techniques especially clustering analysis and association mining in the world of Web services composition was studied in the interest of helping to realize automatic discovery and optimal selection of Web services. The simulation experiments show that clustering analysis can be used to group similar services according to the semantic similarity between different services. The similarity value is a combination of service description similarity, functionality similarity and neighbor relationship similarity. The clustering result can enable service matchmaker to significantly reduce search space, and deploy the discovery of candidate services quickly. On the other hand, if association rule techniques focuses on the analysis of service execution logs and determining which services are likely to be composed together, the utilization of service association relationship can potentially improve the success execution rate of composite Web services.