Predicting School Dropout Among International Students in Türkiye Using Machine Learning: A Case Study of Yemeni Students

International students often encounter a range of challenges that can put their academic progress at risk, sometimes leading to dropouts. This study focuses on identifying the key factors that influence such outcomes, using Yemeni students as a case group to explore broader patterns among international students. A structured survey was conducted with 583 Yemeni students. After removing incomplete responses, 545 valid cases remained. From these, 268 students who were still enrolled were excluded (because students who were still attending school had neither graduated nor dropped out and were outside our scope of study), leaving 277 complete records (128 graduates and 149 dropouts) for the modeling phase. A total of fifteen supervised machine learning algorithms were applied, with training and evaluation carried out using an 80/20 split. Model performance was assessed through common classification metrics such as accuracy, precision, recall, and F1-score. Several important predictors were identified, including academic performance (GPA), proficiency in Turkish, satisfaction with their academic department, financial stability (e.g., access to scholarships, family income), and levels of psychological stress. Among the tested models, XGBoost performed best, achieving 91% accuracy and an F1-score of 0.92 for the dropout class. To illustrate the practical implications of this research, a prototype web application was also developed. Overall, the study demonstrates that machine learning can be a valuable tool for anticipating dropout risks among international students and highlights the importance of early, targeted support in academic, linguistic, financial, and psychological domains.