Exploring Student Minds: A New Way to Spot Depression, Anxiety and Stress

Researchers looked at the mental health of 424 Bangladeshi university students who answered an online survey in July 2024, a time of social and political tension. They found that many students reported strong symptoms: two‑thirds felt depressed, over seventy percent were anxious and more than half experienced stress. Almost half of the participants said they had all three problems at once. To make sense of these numbers, the team used machine learning. They tested eight different models by repeatedly splitting the data into training and testing sets. Support‑vector machines worked best for predicting depression and stress, while a tree‑based model called XGBoost was more accurate for anxiety.The researchers didn’t just let the models run blindly. They applied tools that explain why a model made a certain prediction, like SHAP and LIME. These explanations showed that trouble sleeping, feeling mentally tired, and noticing personal behavior changes were common reasons the models flagged a student as at risk. For anxiety, the model paid more attention to what students studied and how much they used the internet. Gender and the type of university mattered more for depression predictions. The study is not meant to replace doctors or give exact numbers about how many students suffer from mental illness. Instead, it shows that explainable machine learning can help researchers see patterns in student data and decide where to focus future studies. The results are specific to this group of students and should not be taken as evidence for the whole student population.