Stock trend forecasting in turbulent market periods using neuro-fuzzy systems

This paper presents a neuro-fuzzy based methodology to forecast short-term stock trends during turbulent stock market periods. The methodology uses two adaptive neuro-fuzzy inference systems; the controller and the stock market process. The model is based on inverse control theory that simulates the stock market dynamics; enabling 1 day ahead forecasting. The proposed methodology is tested and evaluated using real stock shares data of the New York Stock Exchange. Data demonstrates transactions that occurred during four turbulent market periods: the Black Monday of October 19, 1987, the Russian crisis of 1998, the 11th of September 2001 crisis and the credit crisis of 2008.