The effective development of Civic Education work in colleges and universities can promote the cultivation of high-quality talents. Starting from Internet information technology, this paper constructs a digital platform for Civic Education work, in which Shannon’s theory and balanced power allocation are used to optimize channel transmission and enhance the loading speed of Civic Education resources. The XGBoost algorithm is used to effectively cluster students’ learning behaviors as a way to help Civic Education workers understand students’ learning behaviors in Civic Education. The results show that the average response time is 0.327s, and the maximum throughput is 11849.478 bytes/sec. The prediction accuracy of the XGBoost algorithm for student learning behavior is 4.65%, 9.79%, and 21.79% higher than that of LR, RF, GBM, and Stacking classification models, respectively. 9.79%, 21.41%, and 13.5%, respectively. This shows that the university’s thinking and political education work can be developed through the direction of the digital Internet platform, which in turn provides a new research path for the reform and innovation of thinking and political education work.