Sentiment Analysis of Covid-19 Tweets by Supervised Machine Learning Models

Abstract

The COVID-19 virus's transmissibility has sparked intense debate on social media sites, particularly Twitter. As a result, to employ resources efficiently and effectively, a comprehensive assessment of the situation is crucial. Therefore, COVID-19 tweet sentiment analysis is implemented in this research by employing a supervised machine learning (ML) approach. Data is retrieved from Twitter using the Tweepy API, pre-processed using pre-processing techniques, and sentiment extracted and labelled as positive or negative sentiments using the TextBlob library. Three separate feature extraction techniques are used: Bag-of-words (BoW), Term Frequency-Inverse Document Frequency (TF-IDF) combination with 1-gram, and TF-IDF combination with 2-gram. The sentiment is then analyzed using ML classifiers such as Random Forest (RF) and Support Vector Machine (SVM). For clarity, the dataset is studied further using the deep learning method which is Long Short-Term Memory (LSTM) architecture. The four standard evaluation metrics, Receiver Operating Characteristic (ROC), and Area Under the Curve (AUC) were used to evaluate the performance of the models. The findings show that the RF classifier surpasses all other models with a 0.98 accuracy score when combining 2-gram TF-IDF features. In summary, the model may be used to categorize perspectives and will assist policymakers in making more educated decisions about how to respond to the current pandemic.