A new sentiment analysis model is proposed to classify students' reviews on MOOCs. The model, called BERT-LSTM-CNN (BLC), is based on deep learning methods. It uses pre-trained Bidirectional Encoder Representations from Transformers (BERT) to extract text features, Long Short-Term Memory (LSTM) to capture semantic relations between words and sentence context, and Convolutional Neural Network (CNN) to detect complex local features. The model outperforms existing machine learning and deep learning models in sentiment analysis tasks. Sentiment analysis of students' written feedback is essential for tutors to monitor student behavior and intervene when students are bored or confused. In online learning, especially in MOOCs, tutors have no direct interaction with students, leading to disinterest and dropout. Sentiment analysis helps identify these issues and enables adaptive systems to provide personalized learning paths and interventions. The study shows that deep learning models, such as BERT, LSTM, and CNN, are more effective than traditional lexicon-based models in extracting emotions from text. The proposed model improves the accuracy and robustness of sentiment analysis, which is crucial for enhancing the learning experience in online courses. The research highlights the importance of using machine learning and natural language processing in online education to support students and improve course completion rates.A new sentiment analysis model is proposed to classify students' reviews on MOOCs. The model, called BERT-LSTM-CNN (BLC), is based on deep learning methods. It uses pre-trained Bidirectional Encoder Representations from Transformers (BERT) to extract text features, Long Short-Term Memory (LSTM) to capture semantic relations between words and sentence context, and Convolutional Neural Network (CNN) to detect complex local features. The model outperforms existing machine learning and deep learning models in sentiment analysis tasks. Sentiment analysis of students' written feedback is essential for tutors to monitor student behavior and intervene when students are bored or confused. In online learning, especially in MOOCs, tutors have no direct interaction with students, leading to disinterest and dropout. Sentiment analysis helps identify these issues and enables adaptive systems to provide personalized learning paths and interventions. The study shows that deep learning models, such as BERT, LSTM, and CNN, are more effective than traditional lexicon-based models in extracting emotions from text. The proposed model improves the accuracy and robustness of sentiment analysis, which is crucial for enhancing the learning experience in online courses. The research highlights the importance of using machine learning and natural language processing in online education to support students and improve course completion rates.