Prediction of Volatility Using Monetary Rate and GARCH-LSTM Hybrid Model

Abstract

Predicting volatility is very important for the financial markets as it helps to determine risk and decision-making. Predicting volatilities for such stock indices, which include the Nifty 50, is important for traders, investors, and policymakers. In this study, advanced hybrid models are used to predict the volatility of the Nifty 50 index over intervals of 1, 7, 14, and 21 days. The GJR-GARCH-LSTM and the GARCH-LSTM are two hybrid models that forecast the volatility of the Nifty 50. The effect of including the cash reserve ratio (CRR) in the analysis is also looked at. As the forecast horizon grows, the results show decreased prediction accuracy. The mean squared error (MSE) increased by 0.78% from the 1-day to the 7-day forecast, decreased by 2.63% between the 1-day and 7-day projections, rose by about 55% from the 7-day to the 14-day forecast, and grew by 56% between the 14-day and 21-day projections. The GJR-GARCH-LSTM model had better results compared to the simple GARCH-LSTM hybrid model. The novelty of this study is in building and validating hybrid models, specifically the GJR-GARCH-LSTM, to predict Nifty 50 index volatility and using the CRR as a macroeconomic explanatory variable. Different from current literature, which tends to use hybrid models in a generic sense, this paper adapts the model to the Indian financial environment and shows the additional predictive power of monetary policy determinants such as CRR.