Analyzing COVID-19 Sentiments on Twitter: An Effective Machine Learning Approach

Abstract

The COVID-19 pandemic has brought about a surge in online discussions and social media activity, making it crucial to analyze public sentiment towards the virus and related topics. This thesis focuses on Sentiment Analysis of COVID-19 data on Twitter, employing Multinomial Logistic Regression as the primary classification algorithm. This research explores Sentiment Analysis of COVID-19 data on Twitter using Multinomial Logistic Regression. It constructs a tweet dataset reflecting various sentiments—positive, negative, and neutral. The data undergoes preprocessing, and a Sentiment Analysis model is built, with 70% of data for training and 30% for testing. The model uses Count-Vectorizer, Tf-idf for feature extraction, and Multinomial Logistic Regression to classify tweets. The study achieves state-of-the-art results with a high accuracy of 95.14%, demonstrating the effectiveness of this approach. The results offer valuable insights into public sentiment during crises, aiding in decision-making and communication strategies.