Exploring Naive Bayes for Movie Review Sentiment Classification
Zohaib Hasan, Abhishek Singh, Vishal Paranjape, Saurabh Sharma · International Journal of Innovative Research in Computer and Communication Engineering · 2023
This study investigates the performance of Naive Bayes and Logistic Regression classifiers in sentiment analysis of movie reviews using two feature extraction methods: Bag of Words (BoW) and Term Frequency-Inverse Document Frequency (TF-IDF). We utilized a dataset of 50,000 IMDB reviews, preprocessed through denoising, stop word removal, and stemming. The reviews were then vectorized using BoW and TF-IDF techniques. Our analysis reveals that Logistic Regression outperforms Naive Bayes in terms of accuracy, with Logistic Regression achieving 89.52% accuracy for BoW and 89.23% for TF-IDF, while Naive Bayes obtained 85.01% and 85.74%, respectively. Despite its slightly lower accuracy, Naive Bayes demonstrates a consistent performance with a minimal gap between training and testing accuracies, suggesting robust generalization capabilities. The findings indicate that while Logistic Regression is superior in accuracy, Naive Bayes remains a competitive choice due to its simplicity and consistent performance across different feature extraction methods. This comparison provides valuable insights for selecting appropriate classifiers and feature extraction methods for text classification tasks in sentiment analysis.