Background:The subject of why some people refuse to receive certain vaccines during an ongoing public health crisis is particularly relevant because low vaccination rates will prolong the time it takes for many countries to recover from the coronavirus disease 2019 (COVID-19) pandemic. Thus, the prediction of user opinion who is taking COVID-19 vaccinations plays a significant role. Existing classification methods has the issue of handling larger dataset, and it becomes increased time complexity. Deep Learning (DL) model is solution for analyzing user opinions with increased tweets, and produces better performance. Objective: The major objective of the work is to design a new methodology for the prediction of user opinion using DL. DL is used to predict the evolution of the user opinion to coronavirus vaccination. Findings: Covid-19 Twitter Dataset has been collected from Kaggle. It consists of 2,35,240, 3,20,316, and 4,89,269 tweets for first, second, and third phase. Results of proposed ESRRNN-LSTM classifier is compared with existing methods like Modified Long Short-Term Memory (MLSTM) and Bi-LSTM. Results of classifiers are measured in terms of precision, recall, f-score, and accuracy. Proposed system has the highest results in terms of accuracy of 91.20%, precision of 90.00%, recall of 88.00% and F-Score of 87.00%. The five fold cross validation has been also performed to classifiers. Novelty: this work is resulted in developing an Entropy State-Regularized Recurrent Neural Network-Long Short Term Memory (ESRRNN-LSTM) for vaccination prediction from user opinion tweets. Parameters of classifier are optimized using the entropy function.