The purpose of this paper is to present the conceptual and architectural design of a web enabled service that supports the implementation of data mining techniques. The particular service is addressed mainly to scientists who lack of data mining domain expertise but want to investigate hidden relationships among their experimental or observational data. A novelty of the proposed service relies on the fact that it proposes a common data presentation format which is scalable, robust and extensible, thus allowing scientists from different applications areas to adapt their raw data and use the proposed service. This particular service introduces the most proper data mining technique by investigating several complementary principles like Bayesian networks, neural networks, decision trees, support vector machines and other core algorithms from the field of machine learning and artificial intelligence. The service has been effectively used in a case study which is briefly described.