A machine learning approach for unraveling the influence of air quality awareness on travel behavior

Urbanization has escalated air pollution levels with subsequent health implications. This study explores the potential of awareness about air quality levels on travelers' choices and proposes machine learning models to predict travel mode under exposure to different air quality levels. These models are Random Forest, XGBoost, Naive Bayes (NB), K-Nearest Neighbor, Support Vector Machine (SVM), and Multinomial Logistic Regression (MLR). The models are trained using data from individuals who have an understanding of air quality levels. The trained model is further used to predict travel mode choices when the knowledge of air quality reaches all travelers. Travel modes are aggregated into open/closed modes, private/public modes, and motorized/non-motorized/metro modes to assess the impact of air quality awareness and modal shift. The model evaluation shows that the Random forest (RF) exhibits the highest accuracy and F1 score. The model demonstrates that as air quality worsens, commuters shift their preferences from open modes of transport to closed modes. Similarly, during periods of deteriorating air quality, commuters exhibit a preference for public transportation over private modes. This study emphasizes the crucial role of disseminating air quality information, empowering individuals to make informed travel decisions, and mitigating health risks.