Fuzzy neural network

Fraud is a common phenomenon in business, and according to Section 24 of the Iranian Auditing Standards, it is the fraudulent act of one or more managers, employees, or third parties to derive unfair advantage and any intentional or unlawful conduct. Financial statements are a means of transmitting confidential management information about the financial position of a company to shareholders and other stakeholders. In this paper, by reviewing the literature, 6 indicators of current ratio, debt ratio, inventory turnover ratio, sales growth index, total asset turnover ratio, and capital return ratio as input and detection of financial fraud as output are considered for the fuzzy neural network. The database was compiled for 10 companies in the period from 2010 to 2018 after clearing and normalizing qualitatively between 1 to 5 discrete numbers with very low or very high meanings, respectively. The fuzzy neural network model with 161 nodes, 448 linear parameters, 36 nonlinear parameters, and 64 fuzzy laws with two methods of accuracy approximation of mean squared error and root mean squared error has been set to zero and 0.0000001 respectively. This neural network can be used for prediction.

The especial importance of capital market in countries is undeniable in economic development via effective capital conduct and optimum resources allocation. Investment in capital market requires decision making in new stock exchanges, and accessing information in the case of future status of capital market. Undoubtedly, nowadays most part of capital is exchanged via stock exchange all around the world. National economies are extremely affected by the performance of stock market, high talent and unknown factors affecting stock market, and this causes unreliability in investment. It is clear that unreliable assets are inappropriate and in other side, for those investors who select stock market as a place to invest this asset is inevitable; thus, naturally all investors struggle to reduce unreliability. The present study compares four different models of predicting stock price, namely, Perceptron network, Fuzzy neural network, CART, Decision tree, and Support vector regression in Iran stock market during 2008 - 2012. Research sample includes 81 firms listed on the Tehran Stock Exchange (TSE). The findings compared in the case of five indicates show that for predicting stock price, using CART decision tree, has lower error than other ones. JEL Classifications: C10, C13, C18