Arshiya Fathima, Dr. C. Berin Jones 03 08 Mental health issues including anxiety stress and depression are becoming more common among college students especially those who are under pressure from their studies and careers Early and precise detection of depressed tendencies can facilitate prompt intervention and psychological assistance In this work we use structured survey data to propose a Cat Boostbased machine learning model for predicting depression levels among college students Age gender academic year CGPA and university background are among the demographic variables included in the dataset along with psychological markers obtained from the PHQ9 survey A cleaned dataset of 2028 samples with six categories of depressionNo Depression Minimal Mild Moderate Moderately Severe and Severe Depressionwas used to train and assess the model Because it can handle categorical variables natively and capture intricate nonlinear correlations without requiring a lot of preprocessing Cat Boost was chosen The experimental findings showed an accuracy of more than 92 surpassing conventional machine learning baselines and attaining performance on par with or superior to deep learning systems like FEBiON Extensive analyses such as confusion matrix evaluation and feature importance visualization verify that PHQ9related items have a considerable impact on prediction outcomes The suggested technique provides a dependable and comprehensible framework for automated depression screening assisting academic institutions and mental health specialists in fostering the wellbeing of students Copyright (c) Journal of Science Engineering Technology and Management Science. All rights reserved 2026