Comparasion KNN, Decision Tree and Naïve Bayes for Sentimen Analysis Marketplace Bukalapak

Abstract

The number of platforms currently makes it easier for users to provide reviews, one of which is the Bukalapak application on the Google Play Store. While both negative and positive reviews can influence the value of the app, users can also be affected by the app's sentiment reviews. Therefore it is necessary to carry out sentiment analysis to classify negative and positive reviews. This research uses review data of 1000 reviews and then classifies them using the RapidMiner application using three methods, namely KNN, Naive Bayes, and also the Decision Tree. The results of the KNN method obtained accuracy values of 85.03%, precision of 84.98%, and recall of 100.00%, then for the Naive Bayes method obtained accuracy values of 73.95%, precision of 100.00%, and recall of 69.26%, and for the Decision Tree method obtained 89.12% accuracy value, 88.62% precision, and 100.00% recall. this can prove that the Decision Tree method is superior to the KNN method and also the Naive Bayes method.