olatility forecasting is an important task for those associated with the financial markets, and has occupied the attention of academics and practitioners over the last two decades. This research paper reflects the importance of volatility in option pricing, security valuation and risk management. It investigates the forecasting ability of Feed-Forward Neural Networks (FNN) using backpropagation learning, Recurrent Neural Networks (RNN), and a Generalized Autoregressive Conditional Heteroskedasticity (GARCH) model based on empirical price and historical volatility data. The performance of the three models is studied and compared using data of S&P 500, DJIA, NYSE and NASDAQ indexes. The results obtained from these selected models are anticipated to have significant market directional ability and lower prediction errors.