This study focuses on the innovation of teaching methods for Civic Education in colleges and universities, and provides a structured knowledge framework for teaching by constructing a Civic Knowledge Mapping and integrating course knowledge points. On this basis, a new classroom teaching mode is designed to integrate online and offline teaching resources to enhance student interaction and participation. A knowledge tracking model of key-value memory network (MKVMN) based on multifeature fusion is proposed to accurately track students’ mastery of Civics and Politics knowledge by capturing students’ multi-dimensional learning behavior characteristics. To optimize the recommended path for students’ personalized learning, an improved ant colony algorithm is introduced to generate personalized learning paths based on students’ individual differences. The experimental results show that when the number of learning units is 0-10 (pre-study period), the improved ACO algorithm model does not have obvious advantages for students’ learning, but when the number of learning units reaches 11-50, the difference between the experimental group students’ learning performance and the control group becomes more and more obvious, so it can be seen that the improved ACO algorithm can obviously improve the students’ Civic and Political Science learning performance. In addition, the IACS-PRA algorithm is especially effective in long path recommendation, which finds the optimal personalized recommendation path through a gradual approach to help students learn Civics and Political Science more efficiently, and provides a practical demonstration for the digital transformation of Civics and Political Science education in the new era.