Student Anxiety Assessment Using Machine Learning Techniques

Anxiety among students in Nigeria has become a growing issue, as it affects the academic performance and mental health of students. However, if anxiety is detected early enough in students, necessary steps could be taken to prevent negative effects and complications. Detecting anxiety early among students is difficult. A significant amount of research has been conducted on student anxiety assessment using machine learning (ML), but there has been limited work on integrating traditional survey data with robust machine learning algorithms for real-time applications. This research presents an anxiety prediction model using Machine Learning algorithms, which include Logistic Regression (LR), Artificial Neural Network (ANN), Reduced Error Pruning Tree (RepTree), Classification and Regression Tree (CART), Support Vector Machine (SVM), and XGBoost. The dataset used to train the algorithms consisted of 1,666 responses from students in Nigerian tertiary institutions. The data consisting of students' information related to anxiety, including age, gender, academic performance, study hours per day, financial concern, sleep quality, social support, and physical health, were collected through the use of Google Forms and manually via paper questionnaires. The dataset was validated using a t-test by comparing it with data collected directly from students from Federal University Lokoja and College of Education Okene. The result from the experiment shows that the Artificial Neural Network (ANN) outperformed all the other algorithms by achieving 96.12 percent accuracy. SVM, XGBoost, CART, LR, and RepTree achieved 83.33 percent, 94.61 percent, 60 percent, 64 percent, and 65 percent accuracies, respectively.