Evaluasi Naïve Bayes dan K-Nearest Neighbor dalam Klasifikasi Sentimen Ulasan Produk Skincare MSGLOW di Tokopedia

Abstract

Intisari— Industri skincare di Indonesia mengalami pertumbuhan pesat, terutama dalam segmen perawatan wajah pria. Platform e-commerce seperti Tokopedia menjadi sarana utama bagi konsumen untuk memberikan ulasan produk, sehingga analisis sentimen terhadap ulasan pelanggan menjadi penting dalam memahami persepsi pasar. Penelitian ini mengevaluasi kinerja algoritma Naïve Bayes dan K-Nearest Neighbor (KNN) dalam klasifikasi sentimen ulasan produk skincare di e-commerce. Data dikumpulkan melalui web scraping dari ulasan pelanggan terhadap produk MSGLOW Men Energizer Facial Wash di Tokopedia, menghasilkan 475 data ulasan. Proses pengolahan teks dilakukan dengan metode text mining, termasuk case folding, tokenizing, stop words removal, dan stemming. Analisis dilakukan menggunakan perangkat lunak RapidMiner, dengan klasifikasi sentimen menjadi positif, negatif, dan netral. Evaluasi model dilakukan menggunakan confusion matrix untuk mengukur akurasi. Hasil penelitian menunjukkan bahwa Naïve Bayes memiliki akurasi sebesar 92,16%, lebih unggul dibandingkan KNN yang hanya mencapai 68,82%. Temuan ini menegaskan bahwa pemilihan algoritma yang tepat berperan penting dalam meningkatkan akurasi analisis sentimen di e-commerce. Selain itu, penelitian ini memberikan wawasan bagi industri skincare dalam memahami opini pelanggan dan mengoptimalkan strategi pemasaran berbasis data.

Kata kunci— analisis sentimen, e-commerce, klasifikasi sentimen, Naïve Bayes, K-Nearest Neighbor

Abstract— The skincare industry in Indonesia has experienced rapid growth, particularly in the men's facial care segment. E-commerce platforms such as Tokopedia serve as a primary medium for consumers to share product reviews, making sentiment analysis crucial for understanding market perceptions. This study evaluates the performance of Naïve Bayes and K-Nearest Neighbor (KNN) algorithms in sentiment classification of skincare product reviews in e-commerce. Data was collected through web scraping, obtaining 475 customer reviews of MSGLOW Men Energizer Facial Wash from Tokopedia. Text processing was conducted using text mining techniques, including case folding, tokenizing, stop words removal, and stemming. The analysis was performed using RapidMiner, categorizing sentiments into positive, negative, and neutral. Model evaluation was conducted using the confusion matrix to measure accuracy. The results indicate that Naïve Bayes achieved an accuracy of 92.16%, outperforming KNN, which only reached 68.82%. These findings highlight the importance of selecting the appropriate algorithm to enhance the accuracy of sentiment analysis in e-commerce. Furthermore, this study provides valuable insights for the skincare industry in understanding customer opinions and optimizing data-driven marketing strategies.

Keywords— sentiment analysis, e-commerce, sentiment classification, Naïve Bayes, K-Nearest Neighbor