Self-Organizing Map Method for Fraudulent Financial Data Detection

Abstract

Self-organizing map (SOM) is an unsupervised neural net work method which has properties of both vector quantization and vector projection algorithms, it is an effective tool to detect some outlier of dataset. Since a fraudulent financial information of company is different from other normal data, and it can be regarded as an outlier. Thus, SOM can be applied in identifying the fraudulent financial data. This paper attempts to use the SOM to detect the fraudulent financial data of a company, in which the financial data are chosen from quarterly financial ratios. Here, the financial ratios are composed of six categories: profitability, solvency, growth capability, risk level, operating ability and cash-flow, total of 16 items. The empirical results show that the SOM can effectively divide the financial data into normal and abnormal groups, in which the abnormal group mainly consists of financial fraud ratios. Based on this, we derive the fraudulent financial data.