A Reference Model for Data Mining Web Services

Web services enable to achieve an interoperability between heterogeneous data mining applications through platform and language independent interfaces. As the number of available data mining services increases, it becomes more difficult to find a service that can perform a specific task at hand. Moreover, there may also be no such single data mining service capable of performing the specific task, but a combination of other existing services may provide this capability. Hence, a standardized set of interfaces and data interchange formats between services is necessary to discover suitable services as well as to enable composition of the services into complex workflows. Semantic Web services aiming to tackle the problem of discovering of demanded functionality by grounding semantic Web services to particular WSDL port types or by an extension of the interface description using references to an ontology. In both cases it requires semantically annotated interfaces with well described input parameters and results produced by the services. Hence, a data minig ontology is required to describe semantic Web services performing data mining tasks. The goal of this paper is not to provide an exhaustive specification of the interface defining all kinds of data types and data mining primitives used in communication with the data mining services, but rather propose a model, which can be used by software developers to unify development of the services using common Web standards and technologies. Additionally, the reference model can serve as a template for implementation of WSRF-compliant data mining grid services.