Modelling and Forecasting Four Market Indices: Autoregressive Integrated Moving Average Model versus Artificial Neural Network Model

Abstract

This study model and forecast the stock prices of 4 stock market indices: the FTSE KLCI, Dow Jones Industrial Average, NASDAQ Composite and S&P 500. We utilize two distinct approaches which are time series analysis and artificial intelligence system. We model the data with Autoregressive Integrated Moving Average (ARIMA) and Artificial Neural Network (ANN) applying the time series data of the 4 selected stock market indices. The time frame was five years, starting on 3 January 2012 to 29 December 2017. After the modelling, the outputs are subsequently compared and contrasted in terms of forecast accuracy, such as MAE, MAPE, RMSE and MSE, and the model with the lowest forecast error is sought to be the best-fitted model. We found ANN model to be superior as it outperformed the time series model by generating the lowest forecast error across four datasets.