Purpose: This study aims to first develop a bus passenger demand prediction model based on industrial factors, population, and traffic dataunder multicollinearity. It can help Busan bus operation.Methods: In orderto address the multicollinearity issues, the research mainly considers PCA (Principal Component Analysis), MLR (Multiple Linear Regression), machine learning (GBDT (Gradient Boosted Decision Trees), RF (Random Forest), and deep learning (MLP (Multi-Layer Perceptron), LSTM (Long Short-Term Memory)), and variable selection for predictive modeling.Results and Conclusion: The industrial factors, population and traffic datasignificantly explain the bus passenger demand. The RF provides the best prediction performance.
Jung et al. (Mon,) studied this question.