Analisis Sentimen Ulasan Produk E-Commerce Menggunakan Naive Bayes Dan TF-IDF

Abstract

ABSTRAK
Perkembangan pesat e-commerce menyebabkan meningkatnya jumlah ulasan produk yang dihasilkan oleh konsumen. Ulasan tersebut mengandung opini dan pengalaman pengguna yang dapat dimanfaatkan untuk mengetahui sentimen pelanggan terhadap suatu produk. Namun, besarnya volume data teks membuat proses analisis secara manual menjadi tidak efisien. Oleh karena itu, diperlukan pendekatan otomatis berbasis machine learning untuk melakukan analisis sentimen secara cepat dan akurat. Penelitian ini bertujuan untuk menganalisis sentimen ulasan produk e-commerce menggunakan algoritma Naive Bayes dengan metode pembobotan Term Frequency–Inverse Document Frequency (TF-IDF). Dataset yang digunakan berupa kumpulan ulasan produk berbahasa Indonesia yang telah melalui tahapan praproses teks, seperti case folding, tokenisasi, stopword removal, dan stemming. Selanjutnya, data direpresentasikan menggunakan TF-IDF dan diklasifikasikan ke dalam kelas sentimen positif, negatif, dan netral menggunakan algoritma Naive Bayes. Kinerja model dievaluasi menggunakan metrik evaluasi seperti akurasi, presisi, recall, dan F1-score. Hasil penelitian menunjukkan bahwa kombinasi metode TF-IDF dan Naive Bayes mampu memberikan performa yang baik dalam mengklasifikasikan sentimen ulasan produk e-commerce. Penelitian ini diharapkan dapat memberikan gambaran sentimen konsumen secara otomatis serta menjadi referensi dalam penerapan teknik text mining pada data teks berbahasa Indonesia.
Kata kunci: analisis sentimen, e-commerce, Naive Bayes, TF-IDF, text mining.

ABSTRACT
The rapid growth of e-commerce has led to an increase in the number of product reviews generated by consumers. These reviews contain users’ opinions and experiences that can be utilized to identify customer sentiment toward a product. However, the large volume of textual data makes manual analysis inefficient. Therefore, an automated approach based on machine learning is required to perform sentiment analysis accurately and efficiently. This study aims to analyze the sentiment of e-commerce product reviews using the Naive Bayes algorithm with the Term Frequency–Inverse Document Frequency (TF-IDF) weighting method. The dataset used consists of Indonesian-language product reviews that have undergone text preprocessing stages, including case folding, tokenization, stopword removal, and stemming. Furthermore, the data are represented using TF-IDF and classified into positive, negative, and neutral sentiment classes using the Naive Bayes algorithm. Model performance is evaluated using evaluation metrics such as accuracy, precision, recall, and F1-score. The results show that the combination of TF-IDF and Naive Bayes provides good performance in classifying the sentiment of e-commerce product reviews. This study is expected to provide an automatic overview of consumer sentiment and serve as a reference for the application of text mining techniques on Indonesian-language textual data.
Keywords: sentiment analysis, e-commerce, Naive Bayes, TF-IDF, text mining.