Development Environment of Application Domain-oriented Knowledge Discovery System KDIST

A universal development environment called KDIST, which is used to develop application domain-oriented knowledge discovery systems (KDS), is presented. The KDS is modeled by a number of function modules, each of which is scheduled to acquire from data knowledge related to a specific domain problem. According to the characteristics of the domain problem, a mining scheme is made for the corresponding function module, including data-preprocessing method and mining algorithm. The discovered knowledge in KDS can then be fused into knowledge bases of existing expert systems. Two real-use KDSs developed by KDIST are introduced and proved to be effective on automatic knowledge acquisition and knowledge refinement. Several data mining algorithms are also discussed.