Sentiment Analysis of Twitter Data Using Machine Learning and Deep Learning

EasyChair Preprint 15299

Abstract

In recent years, sentiment analysis has gained significant attention due to the surge in social media and e-commerce platforms. It involves analyzing people's opinions to determine the polarity, beneficial for assessing customer reviews and identifying social trends. Our thesis focuses on a dataset comprising over 29,530 tweets, aiming to discern whether they contain hateful content. Employing machine learning techniques such as Naïve Bayes, Support Vector Machine, Logistic Regression, and Random Forest, we conducted a classification task, evaluating performance through precision, recall, f1-score, and accuracy. Despite minor variations (1-2%) among the models, Random Forest yielded the highest accuracy at 96.24%. The study didn't conclude there; we extended our exploration to deep learning, specifically employing Bidirectional-Long Short-Term Memory. Surprisingly, the deep learning model's accuracy slightly lagged behind machine learning. Consequently, our final determination is that, for our dataset, machine learning outperforms deep learning. In the course of our research, we delved into the challenges and limitations, providing a comprehensive analysis of our work.

Keyphrases: Models CNN, Naïve Bayes, Random Forest, Support Vector Machine, detection, logistic regression

@booklet{EasyChair:15299,
  author    = {Md.Saiful Islam and Mahede Hasan},
  title     = {Sentiment Analysis of Twitter Data Using Machine Learning and Deep Learning},
  howpublished = {EasyChair Preprint 15299},
  year      = {EasyChair, 2024}}