Gated Recurrent Unit Model with Untrained Heteroscedasticity Element in Modelling Forecast of Bursa Malaysia Stock Return Volatility

Abstract

Heteroscedasticity effects are useful for forecasting future stock return volatility. Stock volatility forecasting can provide business insight into the stock market, making it valuable information for investors and traders. Predicting stock volatility is a crucial task and challenging. This study proposes a hybrid model that can predict the future stock volatility value by considering the heteroscedasticity element of the stock price. The proposed model is a combination of Generalized Autoregressive Conditional Heteroskedasticity (GARCH) and a well-known algorithm of Recurrent Neural Network (RNN) which is Gated Recurrent Unit (GRU). This proposed model is referred to as GARCH-GRU model. The proposed model is expected to improve the prediction accuracy by considering the heteroscedasticity element. First, GARCH model is used to find the model estimation. Then, residual obtained from the model is tested using ARCH effect test. This step is crucial in finding any untrained heteroscedasticity element. The hypothesis of the ARCH test yielded a p-value less than 0.05 indicating there is valuable information remained in the residual where it is also known as heteroscedasticity element. Next, the dataset with heteroscedasticity element undergoes modelling step using GRU model. The experimental results revealed that hybrid GARCH-GRU was able to improve the directional forecast accuracy based on Mean Directional Accuracy (MDA) performance measurement. The proposed hybrid GARCH-GRU model has potential to improve the directional forecasting accuracy by 87% based on the MDA scores obtained in single GRU and hybrid GARCH-GRU. This finding proved that GARCH-GRU model is highly reliable in predicting the stock price and able to assist the investors in making decisions on regards to stock price.