Road Accident Black Spot Identification & Risk Assessment Using GIS and Machine Learning

Abstract

Every year, road accidents cause fatalities to thousands of people; hence there is a need to identify dangerous road sections before further accidents happen. Existing techniques for identifying accident clusters focus only on number of accidents that happened without considering risk drivers. In this paper, an integrated GIS–machine learning framework is suggested where geospatial mapping techniques are combined with machine learning algorithms for not only locating the accident clusters but also assessing risk severity of those clusters. Accidents that happened from 2022 to 2025 were collected from Kaggle dataset. The accidents were cleaned and then KDE was applied on the accidents using GIS platform in order to identify risky areas. Four classifiers – SVM, Decision tree, Random forest and XGBoost have been considered and then compared to each other. IGMLF configuration has performed best out of all other classifiers with 98.74% accuracy, 98.41% precision, 98.18% recall, 98.29% F1-score and 98.86% ROCAUC score.