Article Preview
TopIntroduction
Service-Oriented Computing (SOC) is a new computing paradigm that provides a distributed computing infrastructure for both intra- and cross-enterprise application integration and collaboration (Papazoglou, 2003). It has become one of the hottest topics in recent years, receiving much attention from research community and industry. SOC advocates building distributed applications by means of the composition of services, which greatly changes the way that software applications are designed, architected, delivered and consumed (Zhang & Zhang, 2013). Software is now exposed to an open and collaborative environment constituted by services. With the prevalence of SOC, the number of services and their types published on the Internet have been rapidly growing (Chen et al., 2015). The internet of services is being formed (Tan et al., 2010). The number of services and their types being so large makes how to accurately and efficiently discover the desired services become a problem facing many service consumers (Zhang et al., 2016).
Service clustering is considered to be an effective way to facilitate service discovery by organizing services into clusters of similar services to aid the pruning of the query space (Xia et al., 2015; Zhang et al., 2009). In recent years, several kinds of approaches have been proposed for clustering services to improve the performance of service discovery, mainly including functional-attributes-based (FAB) approaches (Dasgupts et al., 2011) and nonfunctional-attributes-based (NFAB) approaches (Chen et al., 2015; Xia et al., 2011; Zhu et al., 2012). In the current work, we focus on FAB approaches and aim at improving the quality of functionally similar clusters. The previous work on FAB approaches has made some great achievements, but there are still the following problems:
- •
Most prior FAB approaches usually designed for a single type of service documents and make use of the information in WSDL (Web Service Description Language) documents or OWL-S (Ontology Web Language for Services) documents to compute service similarity. However, some new types of services such as RESTful services, Web APIs, and Mashups lack of such a WSDL document or OWL-S document to be used;
- •
Most prior FAB approaches use the space vector model (Salton et al., 1975) to reduce the dimensions of the service documents. Only a few approaches take the topic of services into consideration to compute service similarity;
- •
Most prior FAB approaches neglect to utilize the topological information in service usage histories to compute service similarity. However, the service usage histories also include rich information about the similarity of services which is one of the main issues for better clustering performances.
Thus, questions such as “how to organize services into clusters in the absence of WSDL documents or OWL-S documents” and “how to group services into clusters by using information in the topic of services and topology of service usage histories” are natural to ask.