Topological analysis and prediction of pedestrian crash location distribution in an urban district

The study addresses the well-known and critical issue of pedestrian-related collisions. An innovative methodology is applied to an urban area, the Città Studi district in Milan, Italy, which has considerable foot traffic, with the aim of pinpointing the highest risk places. A key ingredient of the proposed methodology is the 2017–2021 crash dataset gathered by the local municipality police. After transforming the road network of the district into a simple graph, we convert crashes into edge and node weights. Three centrality indices, Betweenness, Closeness, and a novel one, Icentr, are used to the resulting weighted network to determine the relevance of each node from a spatial interconnection standpoint. The results reveal that these weights influence the importance of nodes in terms of the topological structure. To counteract the over-dispersed distributions of crashes across the network, the index values are clustered to determine which nodes should be aggregated. A probability density function (pdf) of crash occurrence is created from the new distribution by adding the index values for each cluster over four years. This pdf is then used to retrieve predicted crash data of one year at random. By comparing the data, edge-weighted Icentr outperforms the other centrality indices in predicting crashes in the fifth year.