Comparative Study of Machine Learning Models for Sentiment Analysis

Sentiment analysis plays a crucial role in understanding public opinions on social media platforms like Twitter. This study performs a comparative analysis of three machine learning models—Bernoulli Naïve Bayes (BernoulliNB), Linear Support Vector Classifier (LinearSVC), and Linear Regression—for binary sentiment classification (positive vs. negative) using the Sentiment140 dataset. The dataset, which contains pre-labelled tweets, is preprocessed through text normalization, tokenization, and feature extraction using the TF-IDF vectorization method. The models are evaluated based on accuracy, precision, recall, and F1-score, along with confusion matrices and word clouds for qualitative insights. Our results indicate that model performance varies significantly depending on the feature representation and classification approach. The study highlights the strengths and weaknesses of each model and provides insights into their effectiveness for large-scale Twitter sentiment analysis. This comparative evaluation aims to guide future research in selecting appropriate machine learning models for real-world sentiment analysis applications.