The detection of management fraud is an important issue for reliability of accounting information. This study uses artificial neural network(ANN) and case-based reasoning(CBR) to model for detecting management fraud on financial statements. In prior research enforcement of financial statements by SEC were introduced as a measure for management fraud, improper auditing, or combination of them. The objective of the study is that prediction model of accounting enforcement support audit reviewers to select company to investigate, auditors to asses audit risk, and investors to make decision based on probability of fraudulent financial statements.
Analytical review and discretional accrual models were applied to select original independent variable sets. The candidate input variables supposed to be filtered through statistical analysis represent only the financial information of the companies neither non-financial information nor auditor's character.
The experimental research was designed to have two prediction models, which have different control samples. The common experimental sample was a sub-sample of the population of companies issuing fraudulent financial statement which the SEC had determined the existence of management fraud.
The result is that in both of Model I and Model II prediction performances of ANN and CBR are better then those of statistical methods such as stepwise multivariate discriminant analysis(MDA) and Logit and that ANN model distinguishes between fraudulent and non-fraudulent companies with superior accuracy to CBR consistently.
