Big Data on the Teaching Effectiveness of Civics Courses in Universities Based on Self-Coding Multilayer Perceptron Model Comprehensive Evaluation Research

Practice education is an important part of talent training. In the new era, it is in line with the social development trend to promote practical education in universities in conjunction with the “Internet+” strategy. In this study, a total of 1350 social practice teams from 23 universities were analysed using an autocoder. Forty-eight indicators were recoded to obtain new eight dimensional reconstructed indicators. These new metrics were then fed into a multi-layer perceptron model to generate predictive values for performance evaluation. Finally, an error analysis of the four different algorithms was performed and confirmed that the AE_MLP algorithm used in this study had the lowest error effect. The model now has a recognition error of only 0.1267, which is a 60.21% improvement in accuracy compared to the traditional GLM algorithm, and can to some extent assist university cognitive workers in making certain decisions about the effectiveness of social practice.