Analyzing Crash Factors Using Artificial Intelligence for Curved Rural Roads

Abstract

The popularity of Artificial Intelligence (AI) in solving real-world problems has increased lately. In this research, AI is employed to produce a causal analysis of crash factors on curved roads in rural Pennsylvania. Factors like driver, road condition, environment, and vehicle type are investigated with data from police reports. The knowledge gained from this analysis will assist scholars in the field of AI in their efforts to solve problems that are complicated and time consuming, or otherwise impossible to solve without interventions like these. The AI approach utilized a Bayesian network model with the help of Bayesia Lab software, which provided a user-friendly GUI and efficient algorithms to create efficient models to quantify and correlate these factors to come up with optimal values. This is a quantitative study employing relation-based and descriptive methodologies to answer research questions like what the most significant factors that contribute to crash fatality, crash injury and property damage only (PDO) are, what the relationships between driver behavior, environmental factors, vehicle type, and road factors, in relation to the dependent variables are, what optimizations are possible, and what inferences can be drawn. The findings showed that wet road conditions is the most significant factor for injury and PDO. It also showed that impaired driving is the most significant factor for fatality. The same trends were observed in the optimalizations and inferences in refence to both temporal and spatial parameters, using parameter learning that followed the unsupervised learning.