Artificial neural networks vs. ARIMA-GARCH in stock market prediction : the case of Tunisia and Morocco

Abstract EN

The objective of the present paper is to predict the future evolution of the Moroccan and the Tunisian stock markets using Artificial Neural Networks namely, the Multilayer Perceptron with Back-propagation, and the Auto Regressive Integrated Moving Average with Conditional Heteroskedasticity (ARIMA-GARCH).

Data consisted of daily closing stock prices from 2013 to 2016 (785 observations).

Results showed that artificial neural networks have produced a much lower prediction error compared to ARIMA-GARCH.

It was concluded that ANNs are much more powerful than ARIMA-GARCH.

However, their predictive ability is closely related to how well they are designed