Author ORCID Identifier:
Date of Graduation
7-2026
Document Type
Thesis
Degree Name
Master of Science in Civil Engineering (MSCE)
Degree Level
Graduate
Department
Civil Engineering
Advisor/Mentor
Hernandez, Sarah
Committee Member
Sasidharan, Lekshmi
Second Committee Member
Mitra, Suman
Keywords
traffic Safety, machine learning, campaigns, safe driving behavior
Abstract
Traffic safety campaigns are widely used to promote safer driving behaviors by increasing public awareness of seatbelt use, distracted driving, speeding, impaired driving, and motorcycle safety. However, these campaigns are only effective if their messages are recalled by the target audience. This study investigates the factors influencing traffic safety campaign recall using explainable machine learning. Data from the 2023 and 2024 Arkansas State Police Highway Safety Office Safety Awareness Survey, consisting of 677 valid responses after cleaning the incomplete data, were analyzed for five statewide traffic safety campaigns. Six machine learning algorithms, including Probabilistic Neural Network, Decision Tree, Naïve Bayes, Random Forest, Support Vector Machine, and Extreme Gradient Boosting, were evaluated using the original dataset and three class imbalance treatment techniques: Synthetic Minority Oversampling Technique, bootstrap sampling, and equal size sampling. Model performance was evaluated using weighted accuracy, and Shapley Additive Explanations (SHAP) were used to interpret the best performing model for each campaign. The results showed that the best model and sampling strategy varied across campaigns, with the highest weighted accuracies ranging from 0.589 to 0.905 across the five campaign recall models. Compared with BLR, which achieved weighted accuracies between 0.074 and 0.588, the machine learning models consistently demonstrated superior predictive performance for four of the five campaigns and substantially improved prediction for highly imbalanced datasets. SHAP analysis identified driving frequency, speeding behavior, seatbelt use, perceptions of traffic law enforcement, age, education, and traffic safety attitudes as the most influential factors associated with campaign recall. These findings demonstrate that explainable machine learning can improve understanding of campaign recall and provide practical guidance for developing more effective traffic safety communication strategies.
Citation
Tasnim, J. (2026). Factors Contributing to Traffic Safety Campaign Recall: An Explainable Machine Learning Approach. Graduate Theses and Dissertations Retrieved from https://scholarworks.uark.edu/etd/6430
