Abstract: Calculating and grading the sensitivity of railway passenger groups to price, and implementing the classification of passenger groups on this basis have important reference significance for the establishment of new line ticket prices, the floating calculation of existing line ticket prices, the enrichment of frequent passenger marketing means, the implementation of advance ticket discounts and so on. Taking passengers who had repeatedly taken highspeed railway EMU sleeper as the sample, based on the change of travel behavior after dynamic adjustment of ticket prices, this paper used K-means clustering and BP neural network to identify and evaluate the price sensitivity of each recumbent passenger, and eventually divided the recumbent passenger group into three categories. The results show that this method can better evaluate the passenger price sensitivity, accurately identify the passenger groups with higher price sensitivity, and provide help for railway passenger transportation to achieve the goal of "cutting peak and filling valley" by price means, reduce passenger flow fluctuation and improve the market-oriented operation efficiency.