A comparative study of hybrid and individual models for predicting the Moroccan MASI index: Integrating machine learning and deep learning approaches

Abstract

Forecasting financial market fluctuations is inherently challenging due to its complexity, volatility, and non-linear behavior. This research investigates the predictive accuracy of novel machine learning (ML) approaches for forecasting stock prices. Our approach combines individual ML and deep learning (DL) techniques to predict the daily price of the Moroccan all-share index (MASI). This study introduces novel hybrid models, specifically SVR-XGBoost, MLP-XGBoost, and LSTM-XGBoost. Daily closing price data for the MASI index and sector indices, from 2013 to 2023, is collected. The dataset is used to train and optimize Support Vector Regression (SVR), XGBoost, Multilayer Perceptron (MLP), and Long Short-Term Memory (LSTM) models using the Grid search (GS) algorithm. The performance of these individual models is compared with the hybrid model using standard metrics such as mean absolute error (MAE), mean square error (MSE), mean absolute percentage error (MAPE), and root mean square error (RMSE). In addition, backtesting and bootstrapping interval from the Skforecast library are used. The results demonstrate that the hybrid model achieves the highest accuracy. Moreover, this research holds significant value for investors, financial analysts, and policymakers by refining investment strategies and improving risk management practices.