A new model for black spots identification using Weighted Severity Index

Abstract

In Morocco, road accidents kill about 3,500 people and cause about 100,000 injuries each year. Between 2009 and 2019, the number of accidents has seen an increase of 53.43%. Identification of road black spots is an important task in the process of road safety and plays a vital role in reducing the number of accidents. Indeed, among the various techniques used for treating this issue is the weighted severity index method. It combines accident casualties by weighting each of them with a specific score. For this purpose, we use three ensemble methods, which are capable of attributing importance scores to model features (accident casualties). This paper shows the possibility of combining feature importance tool of XGBoost with Weighted Severity Index method in order to improve identification of accident Black Spots. The analysis of 1584 sections on rural roads in Morocco shows that 173 areas are classified as black spots. Our approach turned out to be efficient in identifying locations with high risk of accidents. In consequence, road sector stakeholders in the country may consider these results with the aim of improving the road safety in the future.