Abstract: In response to the dynamic identification of vessel navigation risks in coastal waters, a real-time vessel risk assessment and group clustering method is proposed. The study utilizes data from the Automatic Identification System (AIS) to calculate vessel collision risk levels in real-time across time and space, constructing a relationship topology based on vessel collision risk and a vessel conflict network. Building upon this foundation and simultaneously considering the magnitude of risk values and distances, a criterion for estimating the number of clusters based on module size is proposed. By employing a spectral clustering method, vessels within the waters are grouped into clusters with tight intra-group conflict associations and sparse inter-group associations, effectively identifying hotspots within local water areas. Results indicate this method can timely and accurately identify risk hotspots in complex navigation environments in coastal waters, aiding maritime regulatory authorities in gaining a more comprehensive understanding of real-time vessel navigation patterns to ensure maritime safety.