Fraud detection at financial transactions by genetic algorithm method

Fraud Audit is considered within the context of financial audit, today. The huge amounts of loss faced by institutions due to fraud and abuse indicate the inadequacy of the approaching the "fraud audit" within the context of financial audit at some points, today. This situation reveals the necessity of fraud audit to exceed the limits of financial audit. Internal control, big data analytics, law and human psychology are the new areas that are needed in fraud audit, today. The huge magnitude of losses faced indicates the importance of forecasting of fraud and abuse before the offence is committed with a new interdisciplinary approach. A fraud forecast modelling is done in this research. The article of K. RamaKalyani and D. Uma Devi (2012) is taken as reference article. In the mentioned article, the authors determine five rules of fraud and detect the fraud transactions by genetic algorithm. In this thesis, the code of the rules that is in Java language in the reference article is changed to Matlab. The code in Matlab is applied to the data set and the results are found in line with the result of the code of the authors in Java. Later, an alternative method is proposed and applied to the data set and new results are found. In the third chapter, the Matlab code is applied to a new, similar and bigger data set and some rules are modified and the fraud transactions are detected. The output of the research shows that some fraud rules may be determined and applied to small or big data sets. This method may assist the auditor to focus on some transactions on the data sets that are detected as red flags. Keywords: Fraud Audit, Fraud Forecast Modelling, Red Flag