Machine Learning Model for Sentiment Analysis of COVID-19 Tweets

Covid-19 pandemic presents unprecedented challenges and enormously affects different aspects of individuals' lives worldwide. The implementation of different prevention measures, the economic and social disruption, and the significant rise in the mortality rate greatly affect the peoples' spectrum of emotions. Sentiment analysis, an important branch of artificial intelligence, uses machine learning techniques to understand public perspectives and gain more insights into how they think and feel. During the pandemic, sentiment analysis increasingly contributes towards making appropriate decisions. This research aims to analyze the public sentiment related to Covid-19 by exploring social perceptions shared on Twitter, one of the most ubiquitous social networks. This goal was achieved by building a machine learning model using a dataset of Covid-19 related English tweets. Different combinations of machine learning classification algorithms (Support Vector Machine (SVM), Random Forest (RF), and XGBoost (XGB)) and feature extraction techniques (Term Frequency-Inverse Document Frequency (TF-IDF) and N-gram) were built and applied to the dataset for binary (positive, negative) and ternary (positive, negative, and neutral) classifications. A comparative study for the performance of the different models was then conducted, and the results concluded that XGB classification algorithm with unigram and bigram for binary classification achieved the highest accuracy of 90%. This sentiment analysis model can assist countries and governments in measuring the impact of the pandemic and the applied prevention measures on people's emotional and mental health and take early actions to reduce their impact or prevent them from becoming severe cases.