Applications of QGIS and machine learning for road crash spot identification

Road intersections or linear segments of roads with frequent road crashes are termed as 'blackspots.' There are many blackspot identification methodologies developed and implemented in global cities with varying results. Cross-country adaptability of such methods is limited due to differences in road design parameters, vehicle modal composition, mixed land uses, and varying driver profiles. The United Nations recognizes road crash fatalities as the eighth most significant cause of death, and detecting their occurrence assumes critical prominence. There is an immediate need to find easy-to-adopt, cost-effective strategies that can assist rapidly urbanizing towns in identifying and predicting crash zones. This study considers Visakhapatnam as a study area for black spot identification using Andhra Pradesh Traffic police crash data between 2014 and 2021 and the Weighted Severity Index (WSI). It studies VMRDA Master Plans between 1988 and 2041 for land use changes and visualizes crash spots in QGIS. Machine Learning Algorithm DBSCAN is used to define crash clusters and validate the latitude and longitude of the crash spots. The results have found that fatalities are high in Gajuwaka and P M Palem Police Station Jurisdictions (PSJs). The study identified seven black spots on National Highway 16 in these two areas by their latitude and longitude. It found that the DBSCAN algorithm is precise and adaptable to analyze the crash spot data. This methodology combining open source QGIS and DBSCAN algorithm is proven applicable for 'black spot' identification in the entire city and for cities with similar traffic problems. Geographic and road profile variations do not limit the scope of this study. Land use variations and their influence on black spot formation are the future scope of this study.