Mohammed Ba-Aoum ^1,2* Mohammed Al-Rezq ¹ Jyotishka Datta ¹ Konstantinos P Triantis ¹

• ¹ Virginia Tech, Blacksburg, Virginia, United States
• ² King Fahd University of Petroleum and Minerals, Dhahran, Saudi Arabia

Self-efficacy plays a pivotal role in determining a student's academic progression and life outcomes. Despite its importance, there is a noticeable gap in the existing literature regarding the evaluation and identification of impactful factors in predicting students' self-efficacy through machine learning, particularly within Muslim societies. Using an empirical dataset collected by a non-profit organization, this research aimed to fill this gap by employing four machine learning algorithms (Decision Tree, Random Forest, XGBoost, and Neural Network) to predict the self-efficacy of secondary school students in Muslim societies. The prediction performance of the models was then compared using root mean square error (RMSE) and r-squared (R 2 ). The predictors incorporated into the models included two demographic variables and 10 distinct factors covering socio-emotional and cognitive traits, regulatory competencies, and specific elements relevant to Muslim contexts, such as religious/spiritual beliefs and collectivist-individualist tendencies. The results showed that Random Forest outperformed the other models in accuracy, as measured by R 2 and RMSE metrics. The findings also emphasized the significance of self-regulation, problem-solving skills, and a sense of belonging as the dominant predictors, collectively contributing to more than half of the model's predictive power. However, variables such as gender, emotion regulation, and the collectivistindividualist orientation did not significantly influence the predictions. Constructs including gratitude, forgiveness, empathy, and meaning-making held moderate predictive value, whereas regional influences and religious/spiritual factors demonstrated a minimal impact. This study not only deepens our understanding of self-efficacy in educational settings but also provides a foundation for data-driven interventions aimed at elevating student performance and well-being. Furthermore, this research highlights the expanding application of machine learning algorithms and data analytics in the domain of education, underlining their capacity to provide data-driven insights into student performance and inform equitable and responsible educational interventions.