Comparison of Naive Bayes Decision Trees and SVM Algorithms for Sentiment Classification of JMO Applications

Abstract

In this study, the researchers found that SVM achieved a precision of 0.75 for negative sentiment and 0.93 for positive sentiment, with recalls of 0.86 and 0.94, and f1-scores of 0.80 and 0.94, and an overall accuracy of 0.88. Naive Bayes showed similar results with a precision of 0.74 for negative and 0.93 for positive, recalls of 0.87 and 0.94, f1-scores of 0.80 and 0.94, and an accuracy of 0.88. Meanwhile, Decision Tree had the lowest precision for negative (0.71) and positive (0.91) sentiment, with recalls of 0.73 and 0.93, f1-scores of 0.72 and 0.92, and an accuracy of 0.85. These findings suggest that SVM and Naive Bayes offer excellent performance in sentiment classification, while Decision Tree, while still effective, performed slightly lower. These results provide valuable guidance in selecting the right algorithm for sentiment analysis on app data. This study compares the effectiveness of three machine learning algorithms—Naive Bayes, Decision Trees, and Support Vector Machine (SVM)—in sentiment classification of JMO apps using review data taken from Google Play Store via web scraping and processed with a Python application. The evaluation is done based on precision, recall, f1-score, and accuracy metrics.