SISTEM ANALISIS SENTIMEN ULASAN PRODUK BERBAHASA INDONESIAMENGGUNAKAN METODE LEXICON DENGAN VISUALISASI INTERAKTIF

Abstract

The rapid growth of e-commerce has led to an increasing number of product reviews submitted by consumers. These reviews serve as an important reference for potential buyers when making purchasing decisions. However, the large volume of data makes manual analysis inefficient. This study aims to develop a sentiment analysis system based on a lexicon approach in the Indonesian language, capable of classifying e-commerce product reviews into positive, negative, and neutral categories, enabling users to upload review data and obtain analysis results through interactive visualizations and downloadable reports. Sentiment classification is performed by matching words in the reviews with a pre-defined sentiment lexicon. To evaluate system performance, testing was conducted using two data split scenarios: 70:30 and 80:20. The results show that the system achieved test accuracies above 79% in both scenarios, with the best performance in the 70:30 split, reaching 80.7% accuracy. These findings indicate that the lexicon-based approach is effective in classifying sentiment in Indonesian-language product reviews and can assist businesses in understanding customer opinions quickly and objectively.