ASSESSING THE RISK OF MANAGEMENT FRAUD THROUGH NEURAL-NETWORK TECHNOLOGY

The purpose of this study is to develop an effective neural network (N N) fraud classification model employing endogenous financial data. Dur ing model training, a NN learns the pattern of input data for a fraud and nonfraud sample. A classification model created from the learned b ehavior pattern is then applied to a test sample. Three models, using different expectation methods to develop data input, act as an investi gation rule to classify financial statement data. NN technology allows the development of pattern aggregation by simultaneously evaluating i solated an analytical procedure (AP) expectations. During the prelimin ary stage of an audit, a financial statement classified as fraudulent signals the auditor to increase substantive testing during fieldwork. Although prior research has examined the effectiveness of preliminary APs as a means of detecting material misstatements, the available evid ence is inconclusive and somewhat ambiguous. The study's results suppo rts future use of NNs as a fraud-risk assessment tool.