Pattern Recognition in Road Safety: Uncovering the Latent Causes of Accidents on Mexico’s Federal Highways

Abstract

Land transportation in Mexico plays a crucial role in ensuring connectivity and facilitating the mobility of both people and commodity. Nevertheless, this sector confronts substantial challenges, predominantly related with road accidents. Understanding the factors that contribute to these accidents is essential to developing and implementing effective safety strategies to reduce their frequency and severity. This research uses two unsupervised methods: latent Dirichlet allocation analysis (LDA) and the K-means algorithm, to identify the underlying factors responsible for road accidents in Mexico. LDA uncovers latent thematic structures in accident reports, revealing patterns in textual descriptions, and K-means identifies groups of accidents that share common attributes. The study period is from the years 2015 and 2019. The results suggest that traffic accidents are significantly influenced by a combination of factors such as driver behavior, road conditions, weather conditions and weather patterns.