Comparing Random Forest with Four Classification Algorithms for Preference Prediction in Air-HSR Intermodal Services

Air and high-speed rail intermodal service (AHIS) is an emerging way of travel service. In this research, we aim to compare the Random Forest algorithm (RF) with other four classification algorithms including the Logistic Regression algorithm (LR), Gaussian Naive Bayes algorithm (GNB), K-Nearest Neighbor algorithm (KNN), Decision Tree algorithm (DT), in predicting travelers’ ticket buying preferences. The research was conducted at Shijiazhuang Zheng Ding International Airport in 2019 to complete a passenger behavior survey. By comparing and analyzing the classification indexes such as accuracy rate and Receiver Operating Characteristic (ROC) Curve, we found that the RF algorithm has the optimal classification prediction performance. Through the RF algorithm, we predict the travelers’ ticket buying preferences of users and make several suggestions for the AHIS operator by changing different personal attributes and travel attributes.