Abstract

Provider By.U is a relatively new and attractive telecommunications service with claims to be the first digital provider in Indonesia. All services are done digitally with the By.U application that offers convenience. Even so not all users are satisfied with the service, there are criticisms and suggestions, one of which is delivered through the By.U app review feature on the Google Play Store. Sentiment analysis is performed to extract information related to provider by.U. The steps taken are scrapping review data, positive and negative labeling, preprocessing data including data cleaning, data normalization, stopword removal and negation handling, sentiment classification using Support Vector Machine (SVM) and TF-IDF as feature extraction. TF-IDF+SVM with 5-Fold Validation produces pretty good accuracy with an average accuracy of 84.7%, precision of 84.9%, recall of 84.7%, and f-measure of 84.8%. The highest accuracy results in fold 2, 86.1%. The effect of TF-IDF on the measurement of model performance is not so great, but it is better.