This paper proposes an efficient attribute approximation algorithm based on partition method. Through the predefined attribute order relationship, the positive region of decision table is decomposed into multiple equivalence classes, combined with the fast sorting method to reduce the computational complexity and realize the efficient attribute approximation. The knowledge structure tree model is constructed, and the core keyword similarity calculation and subtree clustering methods are integrated to obtain the knowledge point aggregation relevance and dynamically aggregate the cross-curricular knowledge points of Civic and Political Education. A collaborative filtering recommendation model based on knowledge association is introduced, combining the TF-IDF algorithm and user similarity measure to optimize the prediction accuracy of resource scores and realize personalized recommendation of resources. The method of this paper is applied to the knowledge association mining of Civics and Political Science courses in colleges and universities, and the optimization of resource recommendation realizes the planning and guidance of students’ learning path. The study shows that the method of this paper can effectively mine the knowledge point association and calculate the aggregated correlation degree, and the aggregated correlation degree of the three main Civic and Political Science courses is more than 0.75, 0.85, and 0.60, respectively. The maximum RMSE value of the recommendation model is 2.03563, the maximum MAE result is 1.50122, and the accuracy of the score prediction is better than the comparison method.
