The innovative design of teaching methods of Civics classes in colleges and universities is an effective measure to help the teaching of Civics classes in colleges and universities. In this paper, the significance of innovative Civics teaching methods is addressed, and a novel path for teaching Civics courses in colleges and universities is proposed. Through feature selection, the K-Means clustering algorithm is optimized, and after optimal selection of the initial clustering center, balanced discriminant function, and combing of the optimization process, the model is applied to the actual innovation of Civics and Political Science teaching, quantitatively assessing the behavioral characteristics of the students and constructing the prediction model to achieve dynamic monitoring of students’ performance. The empirical teaching analysis method is aimed at quantitatively assessing the path proposed in this paper. The clustering results show that consistent with the preset, the model divides students’ learning behavioral characteristics into four categories with clear boundaries and good clustering effect, and most of the student’s grades are concentrated around 60-80 points, according to which the teachers can adjust their teaching strategies in time to improve the teaching effect. Through the control experiment, we know that for the students who have gone through the innovative Civics teaching mode, the mean value of Civics achievement is 71.594, which is 19.329 points higher than the students who have studied in traditional Civics teaching, which is confirmed by the P-value in the t-test of independence which is less than 0.05. It indicates that the innovative teaching method for civics has a significant effect on improving students’ civics achievement.