Sentiment Analysis of Tokopedia Customer Reviews  Using Decision Tree

Dimas Ari Setyawan; Prasetyo Yekti  Utomo; Ridho  Muarief; Deni Nur Fauzi; Nur Hanifa Muslima

doi:10.28926/jdr.v10i1.516

Authors

Dimas Ari Setyawan Politeknik Negeri Madiun
Prasetyo Yekti Utomo Politeknik Negeri Madiun
Ridho Muarief Politeknik Negeri Madiun
Deni Nur Fauzi Politeknik Negeri Madiun
Nur Hanifa Muslima Politeknik Negeri Madiun

DOI:

https://doi.org/10.28926/jdr.v10i1.516

Keywords:

e-commerce, information gain, sentiment analysis, tokopedia

Abstract

This research aims to analyze the sentiment of clothing stores on the Tokopedia marketplace. The dataset comprises the top 13 clothing stores and star ratings ranging from 1 to 5, spanning the period from 2019 to 2024. The Decision Tree algorithm was implemented for the sentiment analysis process, employing three attribute splitting methods: Gini Index, Gain Ratio, and Information Gain. A total of 10,490 reviews were utilized and categorized into three sentiment classes: 4,018 reviews with 4 and 5-star ratings were classified as the positive sentiment class; 3,378 reviews with 3-star ratings as the neutral sentiment class; and 3,094 reviews with 1 and 2-star ratings as the negative sentiment class. The results of the Decision Tree modeling demonstrate that the highest performance was achieved using the Information Gain algorithm, yielding an accuracy rate of 44.23%.