Analisis Sentimen pada Ulasan Aplikasi myIM3 Menggunakan Multinomial Naive Bayes dengan TF-IDF

Abstract

The digital service adoption in Indonesia has emerged as a primary trend to meet the needs of the millennial generation, seeking greater convenience and speed. Amidst this trend, self-service apps like MyIM3 by Indosat Ooredoo Hutchison have become a trusted solution for users to manage their services more efficiently. Sentiment analysis is crucial for understanding user responses to such apps. This study employs the Multinomial Naïve Bayes algorithm with hyperparameter alpha 0.8 and TF-IDF to analyze sen   timent towards user reviews on Google Play Store for MyIM3. The dataset, sourced from Kaggle, consists of 8475 reviews, pre-processed and labeled to 8212 reviews. Model evaluation with an 80:20 split reveals an overall accuracy of 89%, with a precision of 86% for negative (0) and 93% for positive (1) labels. The recall for negative is 95% and positive is 81%. Thus, this research contributes to understanding user perspectives on MyIM3 and provides a basis for enhancing the quality of app-based services.

Keywords: Multinomial Naive Bayes, myIM3, TF-IDF, Sentiment Analysis