An Analysis on Impact Factors of Bus Fuel Consumption Based on Decision Tree Model

This paper focuses on exploring the most sensitive factors affecting bus fuel consumption quantitatively based on the CART (classification and regression tree) model. Firstly, the distribution of fuel consumption data is proved to agree with Gamma distribution after data processing, which introduces the threshold of high-level fuel consumption used in this research. Further analysis on the processed data shows that four factors of road feature like intersection density on the route and six driving behavior factors including rapid acceleration have significant impacts on fuel consumption. Then, the CART model is applied to three sub-datasets of different periods, and the outcomes show that fuel consumption is sensitive to the smoothness of driving during peak hours, the density of bus stops and the frequency of idling at off-peak times. Based on the CART model results, fuel-efficient strategies can be presented quantitatively from aspects of bus line planning, traffic management, and driving behavior training.