Mental health issues among college students are a growing concern, necessitating effective assessment methods to identify individuals at risk and provide timely interventions. In this paper, we propose and evaluate several computational models for mental health assessment based on demographic, academic, and psychological factors. Hence, this paper implemented the Probabilistic Deep Belief Bayesian Network (PDBBN) to classify students' mental health attributes. The proposed PDBBN network computes the probabilistic value of the mental health assessment of the students. With the estimation of the probabilistic model, the extracted features are applied in the Deep Belief Bayesian Network for the classification of student mental health with the Macromedia analysis in college students. The classification is performed with the consideration of information on gender, age, academic performance, social support scores, and self-reported levels of stress, anxiety, and depression, and each model across multiple epochs. Simulation is conducted in comparison with the proposed PDBBN model with the Convolutional Neural Network (CNN), and Deep Neural Network (DNN) models. The results indicate that PDBBN consistently outperforms CNN and DNN in terms of classification accuracy, precision, recall, and F1 score. The simulation analysis of results stated that the proposed PDBBN model achieves a higher classification accuracy of 0.98 which is significantly higher than the CNN and DNN models. Additionally, the proposed PDBBN model expressed that mental health of the students significantly impacts in the academic performance of the students.