Evaluating Sentiment Classification to Specify Polarity by Lexicon-Based and Machine Learning Approaches for COVID-19 Twitter Data Sets

Abstract

As part of data science, sentiment analysis (SA) applied to social media data is a trending research topic. Identifying positive, negative, or neutral opinions or feelings in the text is the attention of sentiment analysis. In the past few years, Social media platforms have become increasingly popular. In this research, natural language processing (NLP) will be employed to extract useful data and information from unstructured text. .The two methods for sentiment analysis covered in this research are the machine-learning method and the lexicon-based method. The paper examines several lexicon approaches to demonstrate the sentiments from Twitter. To increase classification accuracy, it is necessary to use a reliable method with the highest performance. In this study, classifiers such as Support Vector Machine (SVM) and Naive Bayes (NB) were used together with techniques such as TF-IDF (Term Frequency-Inverse Document Frequency) and BOW (Bag of Words). Each algorithm produces a unique outcome. In order to measure the accuracy of classification, metrics such as Precision, Recall, and F-score are considered. This research shows Support Vector Machine (SVM) with TF-IDF is better than other classifiers with an accuracy of 88%.