Hybrid Sentiment Analysis: Majority Voting with Multinomial Naive Bayes and Logistic Regression on IMDB Dataset
Zakaria Zakaria, Andi Sunyoto · 2023
Sentiment analysis plays a critical role in natural language processing as it seeks to identify and categorize the emotions expressed in texts. In this study, we conducted a case analysis using the IMDB dataset to analyze sentiments in movie reviews. We propose an approach that employs a Majority Voting method that merges two effective classification techniques: Multinomial Naive Bayes (MNB) and Logistic Regression (LR). The Majority Voting method we employed considers the decisions made by both MNB and LR, selecting the sentiment that appears most frequently as the outcome. Our objective is to enhance sentiment analysis performance by leveraging the strengths of both methods. We divided the IMDB dataset into an $80 \%$ training set and a $20 \%$ testing set to evaluate the approach. Performance assessment included metrics such as accuracy, precision, and recall. The experimental outcomes significantly improved sentiment analysis performance by combining MNB and LR using Majority Voting. The achieved accuracy was $0.89 \%$ higher compared to using the methods individually. The Majority Voting approach allowed us to integrate the advantages of both classification methods, resulting in more accurate and consistent sentiment predictions in analyzing movie reviews on the IMDB dataset. These findings indicate that the proposed hybrid approach holds great potential in enhancing the performance of sentiment analysis on the IMDB dataset and can be applied to various applications of sentiment analysis.