Comparing the Performance of Machine Learning Algorithms for Emotion Classification on Tweets

Abstract

— The rapid increase in the availability of textual content due to Industry Revolution 4.0 has made sentiment analysis an important area of machine learning research . This study aims to develop a mechanism to identify the hidden emotions in textual content, beyond the three basic sentiments of positive , neutral , and negative . Several machine learning approaches to emotion classification , including Naive Bayes classifiers , Support Vector Machines , Regression , Decision Trees , and Random Forests have been explored . The experiments show that simple linear models can achieve high accuracy (up to 90.5%),suggesting that complex algorithms are not always necessary for effective emotion classification . The performance of the models was evaluated using a variety of metrics , including accuracy , precision , recall , F-score and efficiency . The findings suggest that machine learning approaches can be used to effectively identify emotionsin textual content, even with simple models. This has potential applications in a variety of domains ,such as social media analysis , customer service, and healthcare.