Utilizing machine learning and geographic analysis to improve Post-crash traffic injury management and emergency response systems.
Journal:
International journal of injury control and safety promotion
PMID:
40214190
Abstract
Traffic injuries are a major public health concern globally. This study uses machine learning (ML) and geographic analysis to analyse road traffic fatalities and improve traffic safety in Nakhon Ratchasima Province, Thailand. Data on road traffic fatalities were collected from forensic and hospital records. K-means clustering grouped death locations and identified cluster centres. The Ball Tree algorithm and Google Directions API were used to find the nearest trauma centre hospital from the injury locations. Statistical tests, including chi-square and Kruskal-Wallis, examined relationships between clusters and demographic variables. The analysis identified 181 cases, mostly males (83.43%), with a median age of 37 years. Clustering the death locations into four high-risk areas resulted in a Silhouette Score of 0.94, indicating suitable EMS locations. While no significant correlation was found with demographic variables, distinct patterns were observed in road user types. Testing the prediction performance for the nearest hospital using forty new locations yielded an accuracy, precision, recall, and F1 score of 0.90. These findings emphasize the importance of targeted interventions and resource allocation in traffic injury prevention and emergency response planning, showcasing the potential of ML and geographic analysis in enhancing traffic injury management and emergency response systems.