Analisis Sentimen pada Ulasan Aplikasi Access by KAI Berbahasa Indonesia Menggunakan Word-Embedding dan Classical Machine Learning

Abstrak

Indonesia has a railway application called Access by KAI, released by PT Kereta Api Indonesia (KAI). The public can download and review this application through the Google Play Store. The rating of Access by KAI has declined since 2022, indicating that the application has not met user expectations despite being updated. Reviews on the Google Play Store platform can be analyzed to extract important information, one of which is sentiment. This research conducts sentiment analysis on Access by KAI reviews using word embedding with the Word2Vec model for feature extraction and classical machine learning with Naive Bayes and Logistic Regression for classification algorithms. The Logistic Regression method outperforms Naive Bayes in terms of accuracy and precision with values of 68.83% and 75.49% respectively. However, the Naive Bayes method has an advantage in terms of recall with a value of 45.07%. In this study, reviews of Access by KAI have a predominantly negative sentiment, with 334 out of 400 test data reflecting this sentiment. The words "easy" and "like" are relevant as reasons why reviews have positive sentiment, while the words "application", "pay", and "ticket" are relevant as reasons why reviews have negative sentiment.