Service Oriented Grid Computing Architecture for Distributed Learning Classifier Systems

Grid computing architectures are suitable for solving the challenges in the area of data mining of distributed and complex data. Service oriented grid computing offer synchronous or asynchronous request and response based services between grid environment and end users. Gridclass is a distributed learning classifier system for data mining proposes and is the combination of different isolated tasks, e.g. managing data, executing algorithms, monitoring performance, and publishing results. This paper presents the design of a service oriented architecture to support the Gridclass tasks. Services are represented in three levels based on their functional criteria such as the user level services, learning grid services and basic grid services. The results of an experimental test on the performance of system are presented. The benefits of such approach are object of discussion.