Article

Predicting Student Depression Based on Academic, Lifestyle, and Demographic Factors Using Machine Learning

---

Abstract

The increasing prevalence of depression among students is a critical public health concern, necessitating the development of effective and scalable methods for early identification. This study investigates the efficacy of machine learning models in predicting depression based on a comprehensive set of demographic, academic, and lifestyle factors. Utilizing a dataset of 27,901 student responses, we employed two distinct classification algorithms: Logistic Regression and Random Forest. The data underwent a rigorous preprocessing pipeline, including median and mode imputation for missing values, one-hot encoding for categorical variables, and standardization for continuous features. Both models demonstrated strong predictive capabilities on a held-out test set. The Logistic Regression model achieved an accuracy of 78.80% and a ROC-AUC of 0.8647, while the Random Forest model yielded a slightly higher accuracy of 79.25% with a ROC-AUC of 0.8585. Although both models were highly effective, the Random Forest classifier was identified as the superior model for this application due to its significantly higher recall rate of 84.3%. This metric is paramount in a clinical context, as it indicates a greater ability to correctly identify students who are genuinely at risk, thereby minimizing the number of missed cases. The results confirm that machine learning provides a powerful and reliable tool for proactive mental health screening in educational environments. The successful application of these models has significant practical implications, offering a pathway for universities to implement data-driven systems that flag at-risk students, enabling timely intervention and promoting overall student wellbeing.

Keywords: Depression, Machine Learning, Mental Health, Predictive Modeling, Student Wellbeing, Predictive Modelling

How to Cite:

Endahti, L. & , J., (2025) “Predicting Student Depression Based on Academic, Lifestyle, and Demographic Factors Using Machine Learning”, Artificial Intelligence in Learning 1(3), 195-210. doi: https://doi.org/10.63913/ail.v1i3.31