In this paper, the ideological and political education path is designed according to the spirit of Yan’an. In order to verify the feasibility of the path, as well as to make the student evaluation more effective and objective, the article combines the K-means algorithm and proposes a K-means algorithm based on neighborhood spatial density optimization (NDK-means). The algorithm assists in solving the problem of grading the reliability of student assessment and realizes the labeling of student assessment reliability data. Then, based on the student assessment reliability characteristics and data labeling results, a BP neural network prediction algorithm (CSA-BP) based on crow search improvement is proposed to model the student assessment reliability. The results show that the percentages of freshmen to seniors who are satisfied with the Civic Education proposed in this paper are 85.9%, 82.26%, 76.75%, and 72.45%, respectively, in terms of grade level. Students from urban, rural, and rural areas, in terms of their place of origin, all feel satisfied with the current teaching of Civics and Political Science at 80% or more.