College English major students are taken as the subjects of evaluation, applying learning analysis technology based on learning behaviour data sets and traditional evaluation and comprehensively using the analytic hierarchy process for index weighting and evaluation practice to construct a comprehensive evaluation index system for the core literacy of college English major students. Secondly, this study deeply examines the impact of smart classrooms on the core literacy of college English learners. By applying different regression models, including Ordinary Least Squares (OLS), fixed effects model, and dynamic lagged fixed effects regression, the research results consistently show that the smart English teaching model significantly improves the core literacy of college students. The regression coefficients of all models are between 0.2150 and 0.2818, and they are robust and reliable at a significance level of 1%. In addition, the study explores the role of academic resources as a mediating variable and finds that smart English classrooms improve students' English core literacy by increasing academic resources. Academic resources are confirmed to mediate the positive impact of smart English classrooms on students' English core literacy, producing a mediating effect of 30.78%. Using deep neural networks, this study further explores the complex relationship between core literacy and learning outcomes. Therefore, as an innovative teaching model, the application potential of smart English classrooms in improving students' English core literacy is significant.