Random Fourier Features based approach for Covid-19 Twitter Sentiment Classification using Machine Learning and Deep Learning

The outbreak of Corona has significantly impacted the daily lives of thousands of people. Many individuals turned to social media for guidance and information during this pandemic. However, while social media can be a valuable source of information, it also presents certain drawbacks such as the spread of misinformation. Despite this, social media played an essential role in sharing accurate health information and providing support for individuals struggling with mental health during the pandemic. Social networking sites like Twitter provided a means for individuals to connect and share their experiences during this difficult time. To evaluate the impact of COVID-19, we propose a strategy that utilizes sentiment analysis of tweets from Twitter users. This analysis can help identify the emotions that people are feeling towards COVID-19, such as hope, pessimism, fear, annoyance, sadness, or nervousness. We utilized feature engineering techniques for sentiment analysis to categorize tweets into positive, negative, or neutral categories. We also utilized machine learning models to evaluate the effectiveness of various feature extraction and engineering techniques. In addition, we utilized class imbalance strategies to address the imbalance of emotion classes. Our study compares different feature extraction methods for text data, including statistical methods, word embedding-based methods, kernel feature maps, and hybrid methods. We achieved superior accuracy, precision, f1-score, and recall compared to previous studies when applied to the Covid Senti - A, Covid Senti -B, and Covid Senti -C datasets. Our findings suggest the assessment of public sentiment towards COVID-19 through the analysis of social media data can be a useful resource.