Classification and Evaluation of Driving Behavior Safety Levels: A Driving Simulation Study

Yang, Kui; Haddad, Christelle Al; Yannis, George; Antoniou, Constantinos · 2022 · Crossref

DOI: 10.1109/ojits.2022.3149474

archive: archived pipeline: cataloged verified

Get this paper ↗ (DOI — opens at the source; we link to it, we don't host it)

Summary

This study addresses the critical need for real-time evaluation of driving behavior safety levels, a key factor in road accidents that has not been thoroughly explored in existing literature. While previous research often relied on binary safe/dangerous classifications or focused on specific driving conditions, this paper proposes a comprehensive framework to classify and evaluate driving safety levels in real time. The authors aim to determine the optimal aggregation time interval for high-frequency data, identify the optimal number of safety levels, and develop robust classification models using driving simulation data. The methodology employs a driving simulation experiment conducted with 260 participants (140 with pathological conditions and 120 controls) using a FOERST Driving Simulator. Participants drove on a simulated rural road under varying traffic densities and distraction conditions. The study first determined the optimal data aggregation interval using an improved cross-validation mean square error model based on driver behavior vectors, which identified 1 second as the optimal interval. To classify safety levels, the authors applied three clustering techniques: k-means, hierarchical clustering, and model-based clustering. The optimal number of clusters was determined to be three. Subsequently, supervised machine learning models—specifically support vector machines (SVM), decision trees (DT), and naïve Bayes classifiers—were trained on the clustered data to evaluate safety levels for new observations in real time. The results indicate that the combination of k-means clustering and decision trees provided the highest accuracy for classifying driving behavior into three distinct safety levels. The framework successfully categorized driving behaviors ranging from "normal" to "dangerous," offering a more nuanced assessment than binary approaches. The use of a 1-second aggregation interval effectively captured the necessary information from high-frequency surveillance data without excessive computational burden. The study demonstrates that unsupervised clustering can effectively identify latent safety patterns, which can then be leveraged by supervised classifiers for real-time application. The significance of this work lies in its contribution to Advanced Driver Assistance Systems (ADAS) and real-time traffic safety monitoring. By establishing a validated framework for multi-level safety classification, the study enables more precise warnings and interventions for drivers. The findings suggest that a three-level safety classification is optimal for capturing the spectrum of driving risk, providing a foundation for future research in automated driving safety and real-time risk assessment. This approach overcomes limitations of previous studies that lacked real-time evaluation capabilities or relied on insufficiently granular safety categories.

Provenance

The full processing record for this entry. Every stage of this paper's journey through the pipeline is logged — what ran, with which tool and model, how many attempts it took, and when it last completed.

Summary generated by qwen3.6-27b-prismaquant on 2026-06-26; verification: verified.

Topics

Ranked by relevance to this paper. Hover a topic for its definition.

• exposure measurement

Information type

What kind of knowledge this paper contributes, grouped by family — independent of topic (what it is about) and method (how it was studied).

• Methodological Resource: validation psychometrics, tool software
• Theoretical Contribution: computational model