A Novel Recurrent Neural Network Based Prediction System for Trading

To reduce their exposure to price fluctuations in the markets, traders are increasingly dealing with options and other derivative securities. There is thus a need to address the limitations of traditional parametric pricing methods, which rely on assumptions to capture the complex dynamics of price processes. This paper proposes a novel non-parametric method using a recurrent neural network for estimating the future prices of commodities such as gold and currencies. The price predictions, shown to be accurate and computationally efficient, are used in a hedging system to avoid unnecessary risks. Experiments show that the trading system can, when using the proposed network and strategy, form portfolios yielding a return on investment of nearly 5%.