Bridging Web Mining and Actionable Intelligence: A Decision Support Platform For Threat Detection In Public Security

The management of public security during major events faces a paradigm shift, as the rapid coordination of crowds on the social web transforms digital chatter into real-time physical risks. This article presents a web-based decision support platform for real-time threat detection using Social Web Mining and Natural Language Processing. The system ingests unstructured TikTok data, applies sentiment analysis, and uses semantic topic modeling. It features a dual-layer approach: a macro-view tracking topic growth and sentiment acceleration, and a micro-view for in-depth analysis. A pilot use case based on Brazilian football matches was applied, and the platform successfully identified threat-related topics with viral growth rates, even when sentiment volume varied. By linking semantic patterns and emotional intensity, the system differentiates harmless online chatter from rising hostility. These findings demonstrate that AI-powered monitoring offers law enforcement enhanced situational awareness, enabling timely preventive actions.