Forecasting Volatility with LSTM Techniques

Volatility forecasting is most searched topic in recent times, from past fears there has been tremendous research in this field of finance. This paper aims at forecasting volatility of stock index with high accuracy. The historical volatility was calculated from daily prices using Yang-Zhang method. Deep learning techniques have evolved over the years and have been successfully applied in time series forecasting problems. In this paper LSTM techniques are applied to forecasting volatility 10 days ahead. The performance of the techniques were measured with mean square error and mean absolute error. The performance of LSTM techniques has outperformed Arima, Arfima and Neural network based techniques.