Investigating the factors influencing intercity travel mode choice in urban agglomerations: Insights from a three-phase framework

The accelerated urbanization has led to a continuous increase in transportation demand within urban agglomerations. Accurately predicting intercity travel mode choices and exploring the key influencing factors are crucial for comprehensive regional transportation planning. This study proposes a novel three-phase machine learning (ML)-based framework to ensure modeling accuracy while revealing the mechanisms influencing intercity travel mode choices. Utilizing travel survey data from the Xi’an-Baoji intercity passenger transport corridor within the Guanzhong Plain urban agglomeration of China, we employed three state-of-the-art ML classifiers, named Random Forest (RF), eXtreme Gradient Boosting (XGBoost), and Deep Neural Network (DNN), to predict the travel mode choices. Then, SHAP-based analysis was introduced to reveal the relative importance of different factors in mode choice and examined the interaction effects among the factors. The results demonstrated that the XGBoost model achieved the best predictive performance, particularly exhibiting significant advantages in probability assessment metrics. Travel distance was revealed with the most significant impact on mode choice, which exhibits nonlinear relationships and threshold effects across different modes, reflecting the trends in mode shifts at varying distance levels. Furthermore, compared to long-distance intercity travel, travelers demonstrated a higher sensitivity to the cost-effectiveness of the mode choices. The findings may provide policy and practical implications for transit-oriented strategies in achieving the sustainable development of urban agglomerations.