Exploiting COVID-19 Tweets for Sentiment Analysis Using Chameleon Optimization with Ensemble Learning

In the initial stage of the COVID-19 pandemic, assessing public opinion is the most critical problem to combat the epidemic by enforcing a national lockdown, rendering health services, and presenting quarantine procedures. One popular social media platform that provides public sentiment towards events that happened at the time of the COVID-19 period is Twitter. Since the Twitter post is short and continuously generated. In the domain of Natural Language Processing (NLP), Sentiment Analysis (SA) came up for mining public judgement, while Machine Learning (ML) methods are commonly used for examining sentiments. This research presents a Chameleon Optimization Algorithm with Ensemble Deep Learning Enabled Sentiment Analysis (COAEDL-SA) on COVID-19 Twitter Data. The presented COAEDL-SA technique identifies diverse types of sentiments posted by the user on Twitter. Firstly, data preprocessing is performed to transform the data into a useful format. Next, the COAEDL-SA technique uses the FastText technique for feature extraction. For SA, an ensemble of DL models such as Recurrent Neural Network (RNN), Deep Belief Network (DBN), and Stacked Autoencoder (SAE) is used. Finally, COA is used for the effectual hyperparameter selection of the DL models. The performance assessment of the COAEDL-SA approach takes place on the Twitter dataset. The widespread comparison study highlighted that the COAEDL-SA technique has resulted in improved SA over other DL models.