Impact of Air Pollution on Informed Decision-Making for Choice of a Travel Mode

This study explores the informed decision-making of travelers in Delhi regarding the choice of a travel mode when they have access to air quality information for their trips. For the prediction, different machine learning models are compared with the widely used statistical model (logit model) to understand the insights of these models from behavioral perspectives. The travel modes are categorized into open and closed modes to bifurcate the decisions with respect to exposure to air pollution. The evaluation of the models highlights XGBoost's superior performance in terms of accuracy and F1 score. Results indicate a notable shift in commuter preferences from open to closed transport modes as air quality deteriorates. Additionally, the logit model suggests a preference for open modes for distances up to 2km; however, as distance increases, commuters tend to favor closed modes. By being informed, individuals can make better travel decisions, reducing health risks associated with poor air quality. Logit models can be preferred for the direct interpretability of predictor variables and their coefficients, offering insights into the odds ratios and relationships between predictors and outcomes. Whereas machine learning classification models often prioritize prediction accuracy and adaptability to complex patterns over direct interpretability. This research bridges the gap between machine learning predictions and behavioral insights, emphasizing the potential for informed decision-making in urban transport.