An Architecture For ABEN-KDD- An Agent-based Environment ForKnowledge Discovery In Databases

The ABEN-KDD aims to comprise all phases of KDD process by modeling the knowledge of all specialists involved with them. Based on the principle of knowledge locality, the subsets of highly coupled knowledge existent in the process were identified. As a heterogeneous multi-agent system, it has one agent to each set of these intrinsically related tasks. Some of the agents are more reactive, like ones that execute data mining algorithms, and others have more autonomy, like user interface agents. It was noticed that, in order to reason about data mining knowledge and about the domain knowledge, there must be a general model to which all types of knowledge can be converted. To do this, we use a new model that uses contexts and fuzzy reasoning.