Travel Mode Selecting Prediction Method Based on Passenger Portrait and Random Forest

For airlines, different travel plans should be formulated for passengers in specific target groups accurately. Based on feature data mining for passengers, targeted passenger portraits predict the way of passenger travel and help airlines to achieve accurate marketing. This work presents characteristic data reconstructing method to group passengers with different travel needs. In this paper, the K-means clustering algorithm is used to establish passenger portrait. The clustering algorithm is improved by contour coefficient method. It multi-dimensionally analyses the characteristics of passenger groups. Based on the reconstructed passenger data, a diversified portrait of passengers is constructed for the passengers. The passengers’ characteristics are combined to label the passengers. Passenger portrait is used to predict travel mode selection based on improved Random Forest algorithm, compared with the AdaBoost algorithm. The feature extraction of data mining for passengers with different travel preferences is realized. This paper uses 10 cross-validation to evaluate the algorithm model.