Composite Artificial Intelligence

7Text-Based Analysis of Twitter Data with Machine Learning Models

N. Malathy^*, G. Sharmila, R. Yuvarshini and R. Lavanya

Department of Information Technology, Mepco Schlenk Engineering College, Sivakasi, Tamil Nadu, India

Abstract

Our culture now shares new information on social media as a result of recent technological advancements. Throughout the pandemic, the use of social media services like Twitter has surged. The researchers developed a method to extract the negative, positive, and neutral attitudes from the Twitter Covid-19 dataset. The tweets were categorized using machine learning algorithms for that reason. A machine learning technique was utilized to label the tweets and the established framework is also deployed utilizing the lexicon-based technique to categorize the tweets as good, negative, or neutral. The approach is assessed using several performance metrics, such as accuracy, precision, recall, and F1-Score. Decision trees, random forests, Gaussian nave Bayes, and multi-nominal nave Bayes are examples of models that can be combined rather than built independently. By merging the models, we can improve accuracy using an NLP framework called BERT. It can parse language with “common sense” which is essentially like a human. By using the BERT framework, we get an increased accuracy of 94%. This value indicates that our proposed system can be able to classify the tweets of even a huge volume of datasets.

Keywords: Sentiment analysis, Twitter, tweet classification, ...