Understanding Post-COVID-19 Household Vehicle Ownership Dynamics Through Explainable Machine Learning

Mahbub Hassan, Saikat Sarkar Shraban, Ferdoushi Ahmed, Mohammad Bin Amin, Zoltán Nagy

Understanding household vehicle ownership dynamics in the post-COVID-19 era is critical for designing equitable, resilient, and sustainable transportation policies. This study employs an interpretable machine learning framework to model household vehicle ownership using data from the 2022 National Household Travel Survey (NHTS)—the first nationally representative U.S. dataset collected after the onset of the pandemic. A binary classification task distinguishes between single- and multi-vehicle households, applying an ensemble of algorithms, including Random Forest, XGBoost, Support Vector Machines (SVM), and Naïve Bayes. The Random Forest model achieved the highest predictive accuracy (86.9%). To address the interpretability limitations of conventional machine learning approaches, SHapley Additive exPlanations (SHAP) were applied to extract global feature importance and directionality. Results indicate that the number of drivers, household income, and vehicle age are the most influential predictors of multi-vehicle ownership, while contextual factors such as housing tenure, urbanicity, and household lifecycle stage also exert substantial influence highlighting the spatial and demographic heterogeneity in ownership behavior. Policy implications include the design of equity-sensitive strategies such as targeted mobility subsidies, vehicle scrappage incentives, and rural transit innovations. By integrating explainable artificial intelligence into national-scale transportation modeling, this research bridges the gap between predictive accuracy and interpretability, contributing to adaptive mobility strategies aligned with the United Nations Sustainable Development Goals (SDGs), particularly SDG 11 (Sustainable Cities), SDG 10 (Reduced Inequalities), and SDG 13 (Climate Action).