The capacity to assess and forecast a variety of topics, including commercial requirements, environmental needs, election patterns (polls), governmental needs, etc., may be added to social media as an intelligent platform. This inspired us to start a thorough investigation of public thoughts and opinions on the COVID-19 epidemic on Twitter. The fundamental training data were gathered from tweets. Based on this, we have produced research using ensemble deep learning algorithms to forecast Twitter views more accurately than earlier works that do the same task. An N-gram stacked auto encoder supervised learning technique is used to extract features first. The collected features are subsequently used in a classification and prediction process using an ensemble fusion strategy comprising certain machine learning algorithms, including decision trees (DT), support vector machines (SVM), random forests (RF), and K-nearest neighbors (KNN). Using both mean and mode approaches, all individual findings are combined/fused for a superior forecast. The N-gram stacking encoder we suggest using in combination with an ensemble machine learning strategy surpasses all other known competitive techniques, including bigram auto encoders and unigram auto encoders. The public has a great deal of trust in government policy during the third wave, and they support all measures taken to contain the epidemic, including widespread participation in vaccine programmes.. The study's findings may be summarised by saying that people are getting past their fear of the disease.

Keywords : Omicron Sentiment Analysis, N-Gram, Analysis, Social Media, Omicron, Tweets, Twitter, Big Data, Data Analysis.