<![CDATA[Sentiment analysis is a valuable method for understanding user opinions on digital applications. This study evaluates the performance of the Naive Bayes and K-Nearest Neighbors (KNN) algorithms in classifying sentiments from user reviews of the Duolingo application obtained from the Google Play Store. The dataset consists of 2,000 reviews, comprising 1,000 negative (1-star) and 1,000 positive (5-star) reviews. Preprocessing was conducted in RapidMiner through several stages, including case transformation, tokenization, stopword removal, and stemming, with features represented using TF-IDF. The experimental results show that Naive Bayes achieved an accuracy of 79.96%, recall of 87.76%, precision of 77.21%, and an AUC of 96.40%. Meanwhile, KNN achieved an accuracy of 78.34%, recall of 75.32%, precision of 81.06%, and an AUC of 92.20%. These findings suggest that Naive Bayes outperforms KNN overall, particularly in sensitivity and class separation, while KNN produces more precise positive predictions. Therefore, the choice of algorithm should depend on analysis objectives, whether emphasizing broader sentiment detection or higher precision in positive sentiment classification.]]>
Copyrights © 2025
