Individual Travel Based Transportation Mode Transfer Points Analysis and Identification

摘要

It is very important for urban traffic planning and traffic management to study the traffic modes adopted by individuals and obtain the rules of residents’ travel. The purpose of this paper is to use mobile phone GPS data to identify the transfer points between different transportation modes during individual travel, which is the basis of trip mode segmentation and individual trip mode chain recognition. Existing identification methods mainly focus on methods based on rules and thresholds, which are not universal. Small changes in data distribution will bring large trip mode segmentation errors. To this end, this paper innovatively proposes to use the classification method of ensemble learning to identify individual transition points in the process of travel. At the same time, considering the limitations of traditional evaluation indexes of classification models, this paper innovatively puts forward three model evaluation indexes as supplementary to traditional indexes. The model evaluation indexes shown that Gradient Boosting Decision Tree (GBDT) model has the best performance and can effectively identify transition points.