Designing knowledge discovery based systems in business, finance and accounting with a case-based approach: two case studies
Olivier Curet · 1996
Case based reasoning (CBR) is a recent approach to reasoning which emphasises the relevance of previous experiences in discovering solution strategies. When the emphasis in the application is on learning (rather than simple reuse of existing solutions), generating apposite and similar cases can enhance a novice user's comprehension in tackling unstructured problem domains, for which no standard procedures or coherent rule sets exist. Two recent studies illustrate the potential for learning with a case based system in the domains of top management fraud (TMF) and transfer pricing (TP). In the first example, the implemented system has allowed its users to identify and understand patterns which characterise cases of TMF. In the second example, the system attempts to highlight the subsets of salient features for those companies that have undergone TP audits. Attempts to tackle both these BFA problem domains by conventional analyses have failed or been only partially successful and the solution processes tried have not advanced user understanding. The domains are characterised by the absence of specific rules, knowledge exists mainly in the form of cases, and there are many different types of variables which may be relevant to understanding and problem solving. In both these case studies (TMF and TP), KD in the form of searching and learning through cases has been a promising and novel approach and has provided more legitimate and credible explanations for users. (4 pages)