<![CDATA[Technological developments in this digital era are growing rapidly in various fields, one of which is the field of public transportation. The purpose of this study is to conduct a sentiment analysis of Access by KAI application users on the Google Play Store so that it can be used as a suggestion to improve the quality of the application. This paper uses the Bidirectional Encoding Representations from Transformers (BERT) method with the pretrained IndoBERT model to train the Indonesian dataset. This writing method uses the CRISP-DM method with 6 stages, namely Business Understanding, Data Understanding, Data Preparation, Modelling, Evaluation, and Deployment. The dataset used was 10,000 reviews and after being processed into 9260. The model that was built managed to predict sentiment quite well with a percentage of 85%. However, in neutral sentiment, the number of wrong predictions was more than the number of correct predictions, which was 22 reviews, and the number of wrong predictions, which was 150 reviews. The number of correct predictions for negative sentiment is 2,822 reviews and the number of wrong predictions is 345 reviews. The number of correct predictions for positive sentiment was 234 reviews and the number of wrong predictions was 131 reviews. The model has also been successfully deployed in the form of a website prototype and can strengthen sentiment predictions quite well.]]>
Copyrights © 2024
