Concerning the problem that the Data Envelopment Analysis (DEA) has some shortcomings in showing the differences of weights for the evaluation indexes and sorting and adjusting the efficient decision making units, an improved DEA was proposed. Firstly, the weights of indexes were confirmed by the Analytic Hierarchy Process (AHP) and the preference cone model was built. And then all the decision making units could be sorted in term of the cross efficiency and the parts of the decision making units were adjusted according to the attendance ratio and the ideal decision making units. Finally, the line planning for the Beijing-Shanghai high speed railway was evaluated. The results show that there are four efficient lines among the six lines and two inefficient lines and one of the efficient lines need to be adjusted. The experimental results show that the proposed method can provide theoretical basis for the the adjustment of the train line planning.