(The) hybrid modeling of case-based reasoning for corporate bond rating

Abstract

Case-based reasoning (CBR) is a problem solving technique by re-using past cases and experiences to find a solution to the problems. While other major artificial intelligence techniques rely on making associations along generalized relationships between problem descriptors and conclusions, CBR is able to benefit from utilizing the case specific knowledge of previously experienced, concrete problem situations. A critical issue in CBR is to retrieve not just a similar past case but a usefully similar case to the problem. For this reason, the integration of domain general knowledge into case indexing and retrieving processes is highly recommended in building a CBR system. However, this task is difficult to carry out since such knowledge often cannot be successfully and exhaustively captured and represented. This study proposes a hybrid CBR models using genetic algorithms and inductive learning for an effective knowledge based system. Chapter 3 proposes a hybrid model of CBR using GAs to find an optimal or near optimal weight vector for the attributes of cases in case indexing and retrieving. We apply this weight vector to the matching and ranking procedures of CBR. In Chapter 4, we propose another hybrid approach of CBR using inductive learning technique. While induction compiles past experiences into general knowledge, CBR directly interprets past experiences. Since induction extracts explicit knowledge from the data, CBR can be benefited from an integrated approach. The GA-CBR and the INDUCTION-CBR integration reap the benefits of both systems. Case-based reasoning techniques provide analogical reasoning structures for experience-rich domains while GAs provide case-based reasoning with knowledge through machine learning. We suggest these approaches as a unifying framework to combine general domain knowledge and case specific knowledge. Our proposed approaches are demonstrated by applications to corporate bond rating.