Research on Knowledge Map Construction and Risk Intervention Mining of College Students' Mental Health Characteristics under the Background of Biotechnology

Abstract

Against the backdrop of biotechnology advancement, this study addresses the limitations of traditional college students' mental health assessment methods (such as delayed data collection and limited indicators) by integrating bioinformatics with psychological characteristics. It constructs a college student mental health domain knowledge graph covering five core entity types (mental health care entities, biotechnological characteristics entities, etc.) and seven types of relationships, and develops a multimodal biometric feature fusion model (integrating voice, physiological, and behavioral features) as well as a risk intervention mining model based on graph neural networks. Experimental results show that the multimodal fusion model improves mental health assessment accuracy by 8%-10% compared with traditional single-feature methods; the knowledge graph achieves over 90% accuracy in entity recognition and relationship extraction, effectively identifying potential risk transmission pathways; the intervention mining model enhances intervention effectiveness by approximately 17%. The study identifies key risk factors (academic pressure, employment anxiety, etc.) and biometric predictors (speech rate, HRV, etc.) for college students' mental health, and proposes tiered, group-specific risk intervention strategies. This research enriches college students' mental health assessment methodologies and provides practical support for university mental health education.