Leveraging Artificial Intelligence to Analyze Rural and Urban Traffic Fatalities in Kansas

Abstract: Traffic crashes remain a leading cause of death worldwide, with rural crashes experiencing higher fatality rates than urban crashes. Similarly in Kansas, rural crashes are a significant challenge due to factors like lighting conditions, alcohol consumption, gravel surfaces, and delayed medical responses. Several studies have used traditional statistical models to differentiate between urban and rural crashes but very few have utilized advanced techniques such as Machine Learning. This study investigates the factors contributing to fatal crashes in rural and urban areas in Kansas applying different machine learning models to identify major differences.
Fatal crash data from the Fatality Analysis Reporting System (FARS) provided by the National Highway Traffic Safety Administration (NHTSA) were used in this study. Several machine learning models, such as Logistic Regression, Decision Tree, Random Forest, and Long Short-Term Memory (LSTM) were applied in a classification analysis for crash location as rural or urban. Logistic Regression and LSTM achieved the highest accuracies of 78% and 79%, respectively, and the highest AUC score of 0.84 for both. These two models were particularly effective in identifying rural crashes, with precision and recall scores of 0.80 and 0.88, respectively.
Further analysis using feature importance was conducted to determine the most influencing variables in predicting rural or urban crashes. Both models highlighted a few factors such as lighting conditions, road type, and time of day as some of the most influential in classification. For example, "Dark - Not Lighted" conditions and interstate highways were found to be strong indicators of rural crashes, on the other hand, meanwhile, urban crashes were more likely to occur under daylight conditions and on local municipality roads.
This study demonstrates the value of machine learning models in handling large-scale traffic data, which can be beneficial for informed decision-making in road safety planning. Additionally, by identifying the key contributing factors and differentiating between urban and rural fatal crashes, this study also provides valuable insights for targeted strategies to increase traffic safety for both urban and rural environments.