PEMODELAN SENTIMEN ULASAN PENGGUNA APLIKASI KURURIO DENGAN REGRESI LOGISTIK MENGGUNAKAN PYSPARK DAN SASTRAWI

Abstract

Kururio is a local online transportation application that currently does not have as many users as its competitors. This study aims to understand user perceptions of the Kururio application through sentiment analysis of user reviews on the Google Play Store. The method used is sentiment classification using Logistic Regression on PySpark, with the support of preprocessing using the Sastrawi library. The research stages include review data scraping, data preprocessing (case folding, cleansing, tokenization, stopword removal, stemming), TF-IDF weighting, modeling, evaluation, and k-fold cross-validation. A comparison was made between Logistic Regression, Naive Bayes, Decision Tree, and Random Forest modeling algorithms. The results show that Logistic Regression outperformed the other models, with the optimal train-test split at 70:30. The analysis also revealed that positive reviews were more dominant than negative ones. Therefore, in general, users have a favorable perception of the application, although there are still several aspects that need improvement.