INRECA – A Seamless Integration of Induction and Case- Based Reasoning for Decision Support Tasks

We propose an integrated approach to decision support and diagnostic problems based on topdown induction of decision trees (TDIDT) and case-based reasoning (CBR). This approach has been implemented within the INRECA1 Esprit project. While different ways of integration (Auriol, Manago et al., 1994) and evaluation (Althoff, Auriol et al., 1995; Althoff, 1995a; 1995b) have been carried out in the project, in this paper we want to focus on a deep kind of integration of TDIDT and CBR that we call a seamless one. The INRECA system has been influenced by both the KATE system (Manago, 1989) and the MOLTKE workbench (Althoff, 1992; Pfeifer & Richter, 1993), which includes a.o. the PATDEX CBR system (Althoff & Wess, 1991; Richter & Wess, 1991; Wess, 1993; Althoff, 1993). The main idea is to extend the PATDEX indexing mechanism such that it can process also numeric attributes. For this purpose a k-d tree (Friedman, Bentley & Finkel, 1977) has been introduced in INRECA as one basic structure for indexing and retrieval (Althoff, Wess et al. 1994). While this structure allowed the use of numeric attributes and local similarity measures, it had to be extended to preserve the main qualities of the PATDEX system (Althoff, Wess et al. 1994; Wess, Althoff & Derwand, 1994). Within this paper we will call this extended k-d tree an Inreca tree to avoid confusion with the original one. Based on an Inreca tree both a decision tree and an extended k-d tree can be generated. This allows the use of induced knowledge within a k-d tree to improve the retrieval by the use of generalised knowledge. The more cases are available, the more general knowledge can be extracted and integrated in the Inreca tree for retrieval improvement. In section 2 we will contrast k-d trees and decision trees, in section 3 an overview of the k-d tree improvements is given. Finally, section 4 summarises the use of the seamless integration of TDIDT and CBR in INRECA.