A Comparative Analysis of Machine Learning Models for Long and Short-Term Forecasting of the Egyptian Stock Market: A Focus on EGX30

This study investigates the effectiveness of various machine learning and deep learning models in forecasting stock price trends for the EGX30, the primary index of the Egyptian stock market. Given the unique dynamics of emerging markets, the authors aim to provide a robust predictive framework that assists investors in navigating both short-term (one day to one week) and long-term (one to two months) market movements.

The researchers evaluated nine distinct models, including traditional machine learning (K-Nearest Neighbours, Random Forest, Extra Trees, Decision Tree Regressor, and Gradient Boosting variants like XGBoost, AdaBoost, and LightGBM) and deep learning architectures (LSTM and GRU). To enhance predictive power, the authors performed extensive feature engineering, incorporating technical indicators such as Simple Moving Averages (SMA), Exponential Moving Averages (EMA), and Relative Strength Index (RSI). They also utilized rolling window lag features to capture temporal dependencies and seasonal patterns inherent in financial time-series data.

The study demonstrates that model performance is highly dependent on the prediction horizon. The XGBoost model proved most effective for immediate, one-day forecasts. For longer horizons—ranging from one week to two months—the Gated Recurrent Unit (GRU) network outperformed other architectures. Additionally, the authors found that ensemble techniques significantly improved long-term accuracy, and they noted that the K-Nearest Neighbours (KNN) algorithm showed surprisingly strong performance for long-term predictions, suggesting it remains a viable tool for fintech applications.

Financial markets in developing economies often lack the depth and liquidity of global markets, making standard predictive models less effective. By identifying which specific algorithms excel at different time scales within the Egyptian context, this research provides a practical toolkit for local investors to manage risk and improve decision-making. The study also highlights the value of combining technical indicators with advanced neural networks to better interpret complex market signals.