Abstract:Single-vehicle crashes, particularly those caused by speeding, result in a disproportionately high number of fatalities and serious injuries compared to other types of crashes involving passenger vehicles. This study aims to identify factors that contribute to driver injury severity in single-vehicle crashes using machine learning models and advanced econometric models, namely mixed logit with heterogeneity in means and variances. National Crash data from the Crash Report Sampling System (CRSS) managed by the National Highway Traffic Safety Administration (NHTSA) between 2016 and 2018 were utilized for this study. XGBoost and Random Forest models were employed to identify the most influential variables using SHAP (Shapley Additive Explanations), while a mixed logit model was utilized to model driver injury severity accounting for unobserved heterogeneity in the data collection process. The results revealed a complex interplay of various factors that contribute to driver injury severity in single-vehicle crashes. These factors included driver characteristics such as demographics (male and female drivers, age below 26 years and between 35 and 45 years), driver actions (reckless driving, driving under the influence), restraint usage (lap-shoulder belt usage and unbelted), roadway and traffic characteristics (non-interstate highways, undivided and divided roadways with positive barriers, curved roadways), environmental conditions (clear and daylight conditions), vehicle characteristics (motorcycles, displacement volumes up to 2500 cc and 5,000–10,000 cc, newer vehicles, Chevy and Ford vehicles), crash characteristics (rollover, run-off-road incidents, collisions with trees), temporal characteristics (midnight to 6 AM, 10 AM to 4 PM, 4th quarter of the analysis period: October to December, and the analysis year of 2017). The findings emphasize the significance of driving behavior and roadway design to speeding behavior. These aspects should be given high priority for driver training as well as the design and maintenance of roadways by relevant agencies.

Pay-As-You-Drive Insurance Pricing Model

Good Drivers Pay Less: A Study of Usage-Based Vehicle Insurance Models

Describing Driving Behaviors Based on Decision Optimization Model

Driving Risk Assessment Using Near-Miss Events Based on Panel Poisson Regression and Panel Negative Binomial Regression

Pricing weekly motor insurance drivers' with behavioral and contextual telematics data

Driving risk prevention in usage-based insurance services based on interpretable machine learning and telematics data

Risk Measurement, Risk Entropy, and Autonomous Driving Risk Modeling

A data science approach to risk assessment for automobile insurance policies

Assessing Driving Risk Using Internet of Vehicles Data: An Analysis Based on Generalized Linear Models

Usage-Based Vehicle Insurance: Driving Style Factors of Accident Probability and Severity

Latent Class Analysis of Accident Risks in Usage-Based Insurance: Evidence from Beijing

How risky is distracted driving?

A practical model for pricing optimization in car insurance

Machine learning based car accident risk prediction for usage-based insurance

Unveiling the risks of speeding behavior by investigating the dynamics of driver injury severity through advanced analytics

Estimating Risk Relativity of Driving Records using Generalized Additive Models: A Statistical Approach for Auto Insurance Rate Regulation

Will claim history become a deprecated rating factor? An optimal design method for the real-time road risk model

Semi-autonomous vehicles: Usage-based data evidences of what could be expected from eliminating speed limit violations

Data Mining of Telematics Data: Unveiling the Hidden Patterns in Driving Behaviour

Measuring and Mitigating Biases in Motor Insurance Pricing