Knowledge Graph Construction and Risk Prediction for Early Warning of Individual Extreme Violence

Personal extreme violence is sudden, hidden and highly destructive, and its early identification and early warning have become a key task for public security governance. However, the existing methods based on text analysis and knowledge graph still have obvious shortcomings in risk semantic identification, conceptual system construction and risk assessment interpretability. For this reason, this article is aimed at the early warning of personal extreme violence behaviour, based on the risk ontology library, and then builds a knowledge graph, and puts forward an interpretable risk prediction framework ROKEF. The experimental results show that the method is significantly superior to the representative models of KGGen, GraphRAG and RAKG in terms of key semantic recognition such as high-risk behaviours, extreme speech and psychological abnormalities, and the generated spectrum also performs better in structural integrity and risk correlation. At the same time, the knowledge graph and interpretable forecasting framework based on the ontology library can effectively support the structured representation, quantitative evaluation and traceability analysis of risks. The research results show that this method can provide important theoretical value and practical support for intelligent early warning of individual extreme violence, intelligent public security construction and social security governance.