Using neural networks to detect financial fraud: an applied study on a sample of Iraqi banks listed on the Iraq Stock Exchange

Abstract

Abstract. The research aims to employ (neural networks) in external auditing to predict financial fraud of the research sample banks through the indicators of the Beneish model because the auditing process faces challenges, the most prominent of which is the increase in the size of risks as a result of the test audit (sampling method) and reliance on traditional procedures in data analysis, which creates fundamental risks that are not discovered by the auditor. The reason for this is the deficiency of the planning process and prior assessment of risks, which increases the possibility of the existence of undetected risks. To test the research hypotheses, the researcher relied on the applied approach in the practical aspect, as he used Beneish M-score as one of the analysis tools in the external audit to detect financial fraud in the research sample banks, in addition to using neural networks based on fraud indicators to analyze the financial statements to predict banks that practice fraud. To achieve the research objectives, the researcher relied in the practical aspect on a sample of (10) Iraqi banks listed in the Iraq Stock Exchange for the period (2021-2022). Thus, the researcher concluded that neural networks contribute to predicting financial fraud, as the network accurately distinguished banks that practice financial fraud, indicating the ability of the neural network to distinguish unusual patterns. The researcher recommended the need for audit offices to adopt neural networks as an enhancing tool for external auditing in detecting financial statement fraud..