Predicting Programming Anxiety Among Computer Studies Students Using Automated Machine Learning

Predicting mental health-related challenges, such as programming anxiety, is essential for delivering timely support and improving academic outcomes among students in computer-related disciplines. This study examined the use of Automated Machine Learning (AutoML) to predict programming anxiety levels using a dataset composed of demographic, academic, and behavioral attributes collected from students at a public university in the Philippines. Through the Altair AI Studio Educational 2024.0.0, eight classification algorithms were employed to automatically generate predictive models, which were subsequently evaluated to determine their performance. Among the models produced, logistic regression emerged as the best-performing, achieving the highest accuracy (97.8%), F-measure (98.3%), and recall (99.7%), while also offering advantages in computational efficiency and interpretability. Feature importance analysis identified working student status, previous semester general weighted average, multimodal learning preferences, and access to multiple ICT devices as key predictors of programming anxiety. These results underscore the practical utility of AutoML in educational contexts and highlight its potential for enabling early identification and intervention to support students’ mental well-being and academic performance in computing disciplines.