Educational online platforms have become vital resources for students looking for flexible and accessible ways to acquire knowledge and build skills. These platforms include a wide selection of courses and materials in a variety of areas and fields, appealing to students of all ages and backgrounds. So students and teachers are using these platforms a lot nowadays. Sentiment analysis is a vital component of analyzing user opinion on educational platforms. This study focuses on sentiment analysis of user evaluations from the Google Play Store for educational platforms. This study uses deep learning and transformer algorithms (RoBERTa, CNN, LSTM, BiLSTM, Hybrid CNN-LSTM, Hybrid LSTM-BiLSTM) to extract patterns influencing user satisfaction, preferences, and complaints from a amount of 59391 dataset collected from nine popular educational platform Google Meet, Edureka, Duolingo, Coursera, Programming Hub, BYJU'S, TED, Udemy and 10 Minute School. The findings demonstrate that the two hybrid models perform better than other models in terms of accuracy: hybrid CNN-LSTM and hybrid LSTM-BiLSTM have the accuracies of 93.28% and 93.15% respectively. BiLSTM, LSTM, CNN and RoBERTa all have accuracies of 90.42%, 90.34%, 89.85% and 83.90% respectively