ANALISIS SENTIMEN ULASAN PRODUK E-COMMERCE MENGGUNAKAN NAIVE BAYES

Abstract

The rapid growth of e-commerce in Indonesia has significantly increased the number of customer reviews on marketplace platforms, particularly Tokopedia. These reviews contain valuable customer opinions that can be utilized to evaluate product quality and service performance. However, the large volume of reviews makes manual analysis inefficient and time-consuming. This study aims to implement the Multinomial Naive Bayes algorithm for sentiment analysis of Tokopedia product reviews to automatically classify customer opinions into positive and negative sentiments. The dataset used consists of 500 labeled customer reviews obtained from Kaggle. The research process includes text preprocessing through case folding and cleaning, feature extraction using Term Frequency–Inverse Document Frequency (TF-IDF) with max_features=3000 and ngram_range=(1,2), followed by data splitting into 80% training data and 20% testing data. The classification process is performed using the Multinomial Naive Bayes algorithm. Model performance is evaluated using Accuracy, Precision, Recall, F1-Score, and Confusion Matrix. The experimental results achieved an accuracy of 83%, indicating that the combination of preprocessing techniques, TF-IDF feature extraction, and the Multinomial Naive Bayes algorithm is effective in classifying customer sentiments on Tokopedia product reviews and can support the evaluation of product quality and service performance in e-commerce platforms.