Reducing road accidents remains a paramount concern worldwide, given their profound impact on public safety and well-being. In this paper, we propose a novel methodology for creating a labelled graph data model to predict accident hotspots in urban areas, leveraging accident and road network infrastructure data. The proposed methodology focuses on the creation of a labelled graph where intersections serve as nodes and road segments as arcs by associating accident data with intersections. In order to demonstrate the usability of the proposed methodology, experiments are conducted using simple Machine Learning models to predict accident hotspots, which serve to identify high-risk areas.