Pengembangan Model Klasifikasi Sentimen Dengan Pendekatan Vader dan Algoritma Naive Bayes Terhadap Ulasan Aplikasi Indodax

Abstract

Cryptocurrency trading applications such as Indodax have grown rapidly, the understanding of user sentiment towards the platform is still lacking, so it is interesting to analyze user sentiment towards the platform. To measure sentiment, this research proposes a combined approach of Vader and Naïve Bayes methods. The data used is a collection of user comments on the google play store platform related to user experience using Indodax. The Vader method is used to analyze sentiment directly from the comment text, while Naïve Bayes is adopted to improve accuracy in sentiment classification. The sentiment analysis process involves various steps, starting from data preparation, data pre-processing, labeling of training and testing data and performance evaluation of the Naive Bayes model. At the sentiment analysis stage with the Vader Sentiment method, the positive category obtained the highest percentage of 63.5%, followed by the neutral category at 18.9% and negative at 17.6%. Meanwhile, based on the performance evaluation of the Naïve Bayes model, the accuracy value is 78% while the highest precision value is achieved by the negative sentiment category at 80% and recall in the positive sentiment category at 44%.