Sentiment analysis: A machine learning utilisation for analyzing the sentiments of facebook and twitter posts

Abstract

The article describes a novel sentiment analysis framework for social media platforms, based on a combination of five machine learning (ML) algorithms—Multinomial Naive Bayes, Random Forest Classifier, Gradient Boosting Classifier, K-Nearest Neighbors, and Decision Tree—and three deep learning (DL) algorithms—LSTM, MLP, and CNN. Using comprehensive datasets from Facebook and Twitter, the authors achieved remarkable results, with LSTM demonstrating superior performance, achieving an accuracy of 0.99, and excelling particularly with Facebook data. The authors illustrate the proposed method’s effectiveness through detailed performance metrics, comparing it against existing models. The proposed framework allows for improved accuracy by up to 20.9%, precision by 1.23%, recall by 11.11%, and F1-score by 3.61%. The new method’s effectiveness is confirmed by extensive evaluation on real-world datasets. These new research results enhance sentiment analysis and can be used for better public opinion understanding, business strategy formulation, and decision-making processes. The novelty and scientific contribution lie in integrating diverse algorithms to achieve higher accuracy and more reliable sentiment detection in social media contexts.