Forecasting Fraudulent Financial Reporting Through Artificial Neural Network

In this study, the ability of artificial neural networks (ANN) as a novel method for predicting the likelihood of fraudulent financial reporting of listed companies in Tehran Stock Exchange in a period of 9 years between the years 2006 to 2015 were studied. For this purpose, the information contained in the financial statements and financial ratios and Multilayer Perceptron model, which includes an input layer, hidden layer of visibility software MATLAB, and an output layer is, the likelihood of distorted presentation of the financial report of fraudulent financial reporting through techniques neural network was evaluated. In this regard, the first seven years of information companies, to develop and train the neural network, data validation and verification of the eighth to the ninth year of training, networking and data as test data and test network were designed .Finally, with regard to the results, it was found that the neural network modeling techniques based on neural network integrity is 97.4% and the design and rigorous training, neural networks can be designed with reasonable accuracy the probability to detect and predict fraudulent financial reporting companies.