The objective of this study is to use artificial neural networks for volatility forecasting to enhance the ability of an asset allocation strategy based on the target volatility. The target volatility level is achieved by dynamically allocating between a risky asset and a risk-free cash position. However, a challenge to data-driven approaches is the limited availability of data since periods of high volatility, such as during financial crises, are relatively rare. To resolve this issue, we apply a stability-oriented approach to compare data for the current period to a past set of data for a period of low volatility, providing a much more abundant source of data for comparison. In order to explore the impact of the proposed model, the results of this approach will be compared to different volatility forecast methodologies, such as the volatility index, the historical volatility, the exponentially weighted moving average (EWMA), and the generalized autoregressive conditional heteroskedasticity (GARCH) model. Trading measures are used to evaluate the performance of the models for forecasting volatility. An empirical study of the proposed model is conducted using the Korea Composite Stock Price Index 200 (KOSPI 200) and certificate of deposit interest rates from January, 2006 to February, 2016.