AN ORDINAL MODEL FOR CASE-BASED REASONING IN A CLASSIFICATION TASK

Implementation of a CBR system requires the effective solution of the five interrelated methodological problems: case representation, indexi ng, storage and retrieval, adaptation, and learning and generation. An ordinal model is presented which applies multicriteria decision makin g techniques to the development of case-based reasoning systems. The m odel uses a natural and rather general assumption about the different levels of typicality of different attributes for different diagnostic classes to develop an effective set of algorithms. This approach provi des a flexible tool for case representation, indexing, storage and ret rieval, adaptation, retention, and consistency analysis.