In order to investigate the relationship between horizontal and vertical alignment indicators and accidents of second-class highway, a risk index model based on the back propagation (BP) neural network is constructed to identify the risk change analysis of accident sections and the risk section interval. By collecting historical accident data, we calculate the roadway accident rate under each alignment indicator, and based on the road accident rate, we perform risk ratings of the eight horizontal and vertical alignment indexes. By comprehensively considering the impact of each alignment indicator, we obtain the weight coefficients of each alignment index based on the BP neural network model. By combining the weight coefficient of alignment indicators and risk ratings, we establish the risk