Machine Learning-Based Prediction of Firm Performance using Ownership Structure, Board Diversity, and AI Analytics: Evidence from Jordanian Listed Firms (2015–2024)

الملخص

Purpose - This study combines traditional econometric methods with machine learning (ML) algorithms for predicting firm performance among Jordanian listed firms and ownership structure dimension (institutional, foreign, governmental and ownership concentration) and Board diversity are considered as a predictor and augmented with artificial intelligence driven analytical features Design/methodology/approach - In this study, Panel data of 158 firms listed on the Amman Stock Exchange (ASE) for the period 2015 - 2021 (1,106 firm-year observations) is utilized as the base data set. This is extended to 2024 based on 1,580 observations combination (Monte Carlo simulation) calibrated from historical three variables distribution parameters. The study takes a two-stage analytical framework: (1) system Generalized Method of Moments (GMM) to determine causality and moderation analysis and (2) Random Forest (RF), Extreme Gradient Boosting (XGBoost) and Artificial Neural Networks (ANN) to forecast Return on Assets (ROA) and Tobin's Q (TQ). Findings - GMM results confirm the positive impact of ownership concentration, foreign ownership and government ownership on ROA and positive impact of ownership concentration, foreign ownership and institutional ownership of TQ. Board diversity has a positive moderating effect on the foreign ownership - performance nexus in terms of both measures. ML models have better predictive accuracy (R2 = 0.81-0.89) than the benchmarks of OLS (R2 = 0.23-0.31). XGBoost has the best out-of-sample accuracy, from SHAP analysis, we can see that ownership concentration and firm size are most important in predicting. Originality/value - This is one of the first studies to use a combination of system GMM and ML ensemble methods to examine the relationships between corporate governance and performance in a country in the Middle East emerging market, which offers a new hybrid methodology for studying corporate governance.

الكلمات المفتاحية

• Machine learning
• Firm performance
• Ownership structure
• Board diversity
• GMM
• Random Forest
• XGBoost
• Neural networks
• Jordan
• Amman Stock Exchange

التوثيق

Alrwabdah, F., Alomari, A., AL-Radaideh, I., Almomani, T. M., Alzoubi, R., Rawabdeh, A., & Lok, C. (2026). Machine Learning-Based Prediction of Firm Performance using Ownership Structure, Board Diversity, and AI Analytics: Evidence from Jordanian Listed Firms (2015–2024). Scientific Culture, 12(2.1), 8934-8953. https://doi.org/10.5281/zenodo.19892059