With the advancement of technology and digital transformation, understanding and enhancing students' emotional experiences are key concerns for educators and policymakers. In this study, 2 000 manually labeled review data were first used to train three machine learning models. The trained SVM model was then used to automatically label 76 000 review data. These labeled datasets were subsequently used to train a Convolutional Neural Network (CNN) model. The CNN model converged successfully after 10 iterations and achieved an accuracy of 94.96% on the validation set, significantly outperforming the SVM model. The results show that combining traditional machine learning with deep learning methods can effectively improve the accuracy of sentiment analysis. This provides data support for optimizing educational strategies.