Selected Indian Stock Predictions using a Hybrid ARIMA-GARCH Model

As the stock market time series data (TSD) is highly volatile in nature, accurate prediction of such TSD is a major research problem in time series community. Most of the prediction problems target one-step ahead forecasting, where linear traditional models like auto regressive integrated moving average (ARIMA) or generalized auto regressive conditional heteroscedastic (GARCH) are used. However, if any prediction model is employed for multi-step or N-step ahead prediction, as N increases, two difficulties arise. First, the prediction accuracy decreases and second, the data trend or dynamics are not maintained over the complete prediction horizon. In this paper, a linear hybrid model using ARIMA and GRACH is developed which preserves the data trend and renders good prediction accuracy. Accordingly, the given TSD is decomposed into two different series using a simple moving average (MA) filter. One of them is modeled using ARIMA and the other is modeled using GARCH aptly. The predictions obtained from both the models are then fused to obtain the final model predictions. Indian Stock market data is considered in order to evaluate the accuracy of the proposed model. The performance of this model is compared with traditional models, which reveals that for multi-step ahead prediction, the proposed model outperforms the others in terms of both prediction accuracy and preserving data trend.