Métodos estadísticos para la identificación de patrones de comportamiento de los conductores y la asignación de la responsabilidad aplicada al método de exposición cuasi-inducida

de No, María Almudena Sanjurjo · 2021 · OpenAlex-citations

DOI: 10.20868/upm.thesis.67496

archive: archived pipeline: cataloged verified

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

Summary

This doctoral thesis addresses the critical role of human factors in road safety, noting that human error contributes to 90% of traffic accidents. Despite significant reductions in accident rates in Spain, further improvements require a deeper understanding of disaggregated driver behaviors and accurate risk estimation. The research focuses on two primary objectives: identifying multivariate behavioral patterns of drivers based on gender, age, and accident characteristics, and improving the assignment of responsibility in traffic accidents to enhance the quasi-induced exposure method, which is essential for estimating relative exposure and risk levels. The study utilizes the General Accident Database from the Spanish Traffic General Directorate (DGT), comprising 145,904 drivers involved in two-vehicle accidents on interurban roads between 2004 and 2013. To identify behavioral patterns, the author employs Self-Organizing Maps (SOM), a clustering technique that projects data into reduced-dimensional space, comparing results with K-Means clustering. For responsibility assignment, the research first applies SOM deterministically to analyze a broad set of variables, including offenses, physical defects, and substance use. To address the 9.63% of drivers whose responsibility remained unclassified by the deterministic method, the author develops a probabilistic procedure using Bayes’ theorem and Monte Carlo simulations. This approach treats the responsibility of unclassified drivers as a random variable, propagating uncertainty to the estimation of relative exposure. The findings reveal significant multivariate differences in driver behavior linked to gender and age. Male drivers, particularly the youngest and oldest cohorts, are disproportionately represented in groups committing more offenses or exhibiting unfavorable driving conditions. Differences between male and female drivers intensify when multiple offenses occur jointly. In the responsibility assignment analysis, SOM identified that alcohol and drug consumption influence responsibility, while the impact of physical defects and sudden illness requires further study. The probabilistic model estimated that, with 95% probability, the relative exposure of male drivers compared to female drivers lies between 2.395 and 2.418. Crucially, the deterministic SOM results yielded higher values than the probabilistic distribution, indicating that deterministic methods may overestimate male relative exposure and consequently underestimate their relative risk levels. The significance of this work lies in its methodological contributions to road safety research. By using SOM, the study provides a comprehensive, multivariate analysis of driver behavior and responsibility, surpassing traditional univariate or bivariate approaches. The introduction of a probabilistic framework for responsibility assignment is the first of its kind for the quasi-induced exposure method, offering a more precise estimation of risk levels by accounting for uncertainty. These tools can assist regulatory bodies in allocating resources more effectively and designing targeted safety measures, such as education and awareness campaigns, for specific driver groups identified as high-risk.

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_with_issues.

Topics

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

• induced exposure
• sex gender
• exposure measurement
• causation analyses
• incidence prevalence
• cultural cross national

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).

• Empirical Findings: crash risk outcomes, observational prevalence
• Theoretical Contribution: computational model