Post-pandemic college students face intensified mental health challenges, while traditional mental health education (MHE) struggles to foster long-term resilience. This study addressed this gap by integrating positive psychology with a backpropagation neural network (BPNN), using 900 valid samples from three universities in North China. A scientifically validated questionnaire (Cronbach's α=0.89, Comparative Fit Index=0.92) provided input data, with standardized psychological test scores as model outputs. Results showed the BPNN achieved over 94% prediction accuracy (variance ≤0.05, F(1,899)=427.3, p<0.001) and good computational efficiency (no failures in six training iterations). Positive psychology-based MHE interventions reduced students' stress scores by up to 25%, particularly for those with high initial stress. This framework offers a proactive tool for MHE, though future research should expand sample diversity and incorporate physiological biomarkers to enhance generalizability.