With the advancement in information technology and social media, there is a need for a patient sentiment-based healthcare system (HCS) to gauge its actual performance. Mining patient sentiment can provide valuable information about the strengths and weaknesses of these HCS and also help in their improvement. Sentiment analysis can help transform existing HCS into better and more efficient systems. Traditional sentiment analysis models are inefficient at capturing contextual information and high-dimensional data. This study aims to develop a classification model that allows the administration to evaluate the effectiveness of the healthcare system. This article proposes a hybrid deep neural network model, DeepCLNet, using the whale optimization technique (WOT) to enhance sentiment classification accuracy and fine-grain emotion analysis of patients. This model incorporates an adaptive feature weighting strategy that dynamically adjusts the weights and provides contextual sentiment refinement to deal with medical text. Furthermore, to enhance the capabilities of the model, this study proposes another hybrid model, BEDeepCLNet, by replacing the DeepCLNet embedding layer with BERT embeddings. The proposed models achieved state-of-the-art accuracies of 96% and 98%, respectively. During the experiments, it was observed that the proposed model performed better than other existing deep learning models applied in this domain. We also compared the proposed model's performance with the best NLP transformers like BERT, RoBERTa, XLNet, and SWIN, etc. The proposed model utilizes fewer computing resources while offering better performance than the transformers.

Analysis Deep Learning Whale Optimization Algorithm Healthcare services Natural Language Processing Adaptive Feature Weighting Contextual Sentiment Refinement