From Service Identification to Service Selection: An Interleaved Perspective

Business process implementation can be fastened by identifying component services that can be used to implement one or more process tasks and by selecting them from a repository of already implemented services. In this paper, we provide an iterative procedure to address this issue, by combining the two macro-phases of service identification and service selection. Starting from a workflow-based specification of the business process, service identification is firstly executed. The result of this phase is a decomposition tree, where basic process tasks are progressively organized into sub-processes (the candidate services) by applying an agglomerative clustering algorithm, based on cohesion and coupling metrics. Within the decomposition tree, a set of candidate services that minimize the coupling/cohesion ratio for the overall process is chosen. The service selection phase works on this decomposition and looks for available services. If the service selection phase fails for some candidate services, a revised set of candidate services is selected by leveraging on the decomposition tree.