What type of study is this?

This is a Quantitative Study study.

October 22, 2025Open Access

Sentiment Analysis of Transjakarta App Reviews Using the Naive Bayes Algorithm

Key Points

The analysis achieved a model accuracy of 78%, revealing insights into user satisfaction with the Transjakarta app.
Precision for positive sentiment reached 82%, while negative sentiment had a lower precision of 67%, showing distinct user experiences.
The study utilized the Naive Bayes algorithm alongside TF-IDF weighting and SMOTE oversampling to enhance sentiment classification.
The findings highlight key positive and negative feedback words from user reviews, guiding potential improvements for the app.

Abstract

In utilizing digital technology in the transportation sector, Transjakarta has introduced a mobile-based application to facilitate public mobility. The number of Transjakarta passengers has increased significantly, making it important to know whether users are satisfied with the application or not. This study aims to classify sentiments and identify the aspects and issues that frequently arise in user reviews of the Transjakarta app using the Naive Bayes algorithm. This study employs the CRISP-DM methodology. Analysis was conducted by scraping 1,000 Google Play Store review data based on MOST RELEVANT, following preprocessing, TextBlob data labeling, TF-IDF weighting, and oversampling (SMOTE) methods. The implementation of the Naive Bayes algorithm with an 80:20 resulted in 684 positive data and 315 negative data, yielding a model accuracy of 78%. For the positive sentiment class, the precision was 82%, recall was 85%, and the F1-score was 84%. For the negative sentiment class, the precision was 67%, recall was 62%, and the F1-score was 65%. Based on the visualization, the words that frequently appear in positive reviews are “bagus”, “lengkap”, and “mudah”, and negative reviews are “ribet”, “susah”, and “error”.

Read Full Paperexternally

Ask AI

Helpful

Bookmark

View Full Paper