Abstract
Depression is more than just feeling sad. It is a severe and multifaceted mental health condition that impacts millions of individuals around the globe. Regrettably, it can even be more prevalent in university students of underdeveloped and developing countries like Bangladesh because of academic pressure, family and societal expectations, financial limitations, stigmatized social and cultural norms, unemployment concerns, lack of mental health awareness, etc. Each of these factors can play a significant role in leading someone towards depression, with their impact varying from person to person. This research, along with detecting depression and gaining insights into the reasons behind the prevalence of depression in this specific population, also focuses on providing simple yet important and tailored recommendations to those who need them. To achieve these objectives, a survey was meticulously designed in collaboration with psychologists, counselors, and therapists. Seven machine learning models, including Support Virtual Machine (SVM), K-Nearest Neighbor (K-NN), Gaussian Naive Bayes (GNB), Decision Tree (DT), Random Forest Classifier (RFC), Artificial Neural Network (ANN), and Gradient Boosting (GB), were trained and tested using the collected data (n = 750) to identify the most effective method for predicting depression. After rigorous analysis, Random Forest emerged as the best-performing algorithm, exhibiting remarkable accuracy (87%), precision (78%), recall (95%), and f1-score (86%). This research mainly strives to identify the initial signs of depressive symptoms among Bangladeshi university-going students and facilitate timely and targeted interventions for the affected individuals. By doing so, it ultimately aims to contribute to building a brighter, healthier, and more resilient educational environment in the country.