Harnessing Machine Learning for Sentiment Analysis: A Comparative Study of Classification Models on Discord User Reviews

This research conducts a comparative analysis of machine learning models to gauge user sentiment in Google Play Store reviews of Discord over the past year. As Discord broadens its scope beyond gaming to become a versatile communication tool, understanding user emotions is crucial, particularly in dynamic markets like Indonesia. We evaluated the performance of five classification models-Naive Bayes, Support Vector Machine (SVM), Random Forest, Neural Network, and AdaBoost-using metrics like accuracy, precision, recall, and F1 score. Our findings reveal that SVM and Random Forest lead the pack, offering superior accuracy and robustness in sentiment analysis, while Neural Networks also perform well but fall slightly short. Naive Bayes and AdaBoost deliver reliable results but are less efficient. This study uncovers key sentimental trends: positive reviews often praise features and performance, while negative feedback highlights issues with login and verification. Ultimately, our research underscores the critical role of choosing the right model to enhance user experience and inform strategic improvements on platforms like Discord. Future work should explore real-time sentiment analysis to boost user engagement and satisfaction.