Educational evaluation is the main content of comprehensive evaluation in undergraduate schools, which involves many indicators, including teaching conditions, teaching effects, students’ learning interest and knowledge conversion rate. Therefore, comprehensive analysis is needed with the help of intelligent analysis methods. This paper puts forward the capital index of education evaluation, and makes logical construction and analysis of it. Either way, this content has become the focus of research. The process of teaching evaluation is a dynamic process, and the data involved are multi-index data. Moreover, the comprehensive activities of students, teachers and third-party evaluation are realized, and the design content is relatively large, so it is necessary to make logical judgment with the help of intelligent analysis methods. Next, a convolution neural network assessment model building method is developed using neural network theory. The outcomes of this model creation are then thoroughly examined and assessed. According to the results of the MATLAB simulations, when comparing the convolution neural network and the traditional particle swarm algorithm for assessing the accuracy and time required to evaluate the factors influencing model construction, the convolution neural network comes out on top under specific evaluation criteria.