Identification of Finite-Mixture of ARIMA-GARCH Model for forecasting NIFTY50

Abstract

Forecasting of stock prices and indices using different time-series models always poses a challenge for selection of appropriate model based on its predictive efficacy. It is observed in past that people get very high returns on one time line and make loss on the other time line. Thus, it is difficult for any trader, investor or asset manager to create portfolio value for sustainable growth in portfolio value in today’s scenario. So, here the purpose of this study is to test and suggest the most fitted time-series model. For that, it is required to confirm the normality and more importantly, whether data is stationary or non-stationary. Data were collected for the NIFTY50 indices of India from NSE India. In this study, various ARIMA (p,d,q) models were tested and best-fit model was selected using Box-Jenkins methodology. This selected model was compared with GARCH (1,1) and ARIMA (p, d, q) - GARCH (1, 1) to yield the best-fit model for forecasting of indices. At last, 10 points forecasting were shown using best model of prediction for NIFTY50. Thus, traders and asset managers can leverage the benefits using the suggested model for accurate prediction of indices for the purpose of trading gains.