Analysis on Movie Reviews: A Comparative Study
Muskan Soni, Namrata Kumari, Pardeep Singh, Vikas Kashtariya · 2024
Sentiment analysis, a crucial task in natural language processing (NLP), involves classifying text based on the sentiment expressed, such as positive, negative, or neutral. This study compares the effectiveness of traditional machine learning models—Support Vector Machines (SVM), Logistic Regression, Naive Bayes, and Decision Trees—with the deep learning model BERT, applied to sentiment analysis of movie reviews. The IMDb dataset, consisting of 50,000 reviews, was used to train and evaluate these models. Preprocessing steps, including padding and feature extraction using TF-IDF, were applied to prepare the data for analysis. Feature scaling was particularly important in enhancing the performance of Logistic Regression. The results show that while traditional models like Logistic Regression and SVM performed well, with accuracies of 89.19% and 88.19%, respectively, BERT outperformed them, achieving an accuracy of 93.31%. The study highlights the superiority of deep learning models in capturing the contextual meaning of text, making them more effective for complex sentiment classification tasks. This research provides valuable insights into the practical applications of these models, showing that BERT offers better performance in large-scale sentiment analysis compared to traditional approaches.