Artificial Intelligence for Financial Market Forecasting: Comparative Assessment of ARIMA, ANN, and Hybrid ARIMA–ANN Approaches on Morocco and South Korea

Résumé

Over the past few years, hybrid modeling approaches have attracted growing interest in financial forecasting due to their ability to capture both linear and nonlinear dynamics. This study proposes a data-driven hybrid model combining Autoregressive Integrated Moving Average (ARIMA) and Artificial Neural Networks (ANN) to forecast stock market indices in Morocco (MSI20) and South Korea (KOSPI).
The empirical results show that the ARIMA model exhibits weak predictive performance, characterized by high error rates and very low explanatory power. The ANN model significantly improves forecasting accuracy, particularly for the Moroccan market, although it remains limited in highly volatile environments. In contrast, the hybrid ARIMA–ANN model clearly outperforms both standalone models, achieving substantial error reductions and high coefficients of determination (R² = 0.888 for MSI20 and 0.975 for KOSPI).
These findings confirm that hybrid ARIMA–ANN models provide a robust and reliable framework for financial market forecasting by effectively capturing both linear and nonlinear structures, with superior performance observed in more liquid and stable markets such as South Korea.