Exploring and Improving Upon the Hedging Performance of Neural Networks for the Individual Investor

Abstract

This thesis explores the performance of artificial neural networks in hedging tasks, specifically for the individual/everyday investor with rebalancing constraints. It finds that a basic hedging ANN may only outperform traditional hedging strategies for investors with higher levels of risk aversion. The paper goes on to suggest that including realized volatility in the feature set of hedging ANNs improves their performance and thus makes them useful for a broader range of risk aversion levels.