under the background of energy conservation, emission reduction, green transportation and transportation strengthening national power, taking 18 Railway Group Companies in China as the research object, based on the theory of undesired output DEA model, Tobit model and bootstrap model, a regression model based on Railway carbon emission efficiency is established. The objective function of the model is to maximize the expected output and minimize the unexpected output. The model is solved by linear programming and other theoretical methods. The research results show that the carbon emission efficiency of Railway Group Corporation shows regional differences, and the Gini coefficient decreases from 0.27 to 0.14, indicating regional difference decreases year by year. After 2010, the number of DEA effective companies increased from three to five, namely: Taiyuan Railway Group Company, Jinan Railway Group Company, Guangzhou Railway Group Company, Hohhot Railway Group Company and Lanzhou Railway Group Company. Through the second stage of regression, we can get the following conclusions: in the current transportation network, the competition effect between high-speed railway, civil aviation and highway is greater than its cooperation effect; the efficiency of high-speed railway and carbon emission presents an "inverted U-shaped", that is, the scale of high-speed railway has an efficient scale area which can promote carbon emission efficiency.