Analisis Faktor Kepuasan Pelanggan Produk Skincare di Tiktok Shop Menggunakan Machine Learning dan Shap

Selma Mulkya Nisa (Universitas Bengkulu)
Yudi Setiawan (Universitas Bengkulu)
Endina Putri Purwandari (Universitas Bengkulu)

Jurnal Sains dan Teknologi
Vol. 5 No. 4 (2026): Agustus 2026

Customer reviews of skincare products sold through TikTok Shop provide valuable information for evaluating customer satisfaction and understanding consumer preferences. As the volume of reviews continues to grow, manual analysis becomes increasingly inefficient, making machine learning-based sentiment analysis a more effective approach. This study applies the Random Forest algorithm to classify customer sentiment and employs Shapley Additive Explanations (SHAP) to interpret the factors influencing the model's predictions. The study utilized a dataset of 663 customer reviews collected from TikTok Shop. The analysis process consisted of text preprocessing, TF-IDF feature extraction, sentiment labeling, Random Forest model development, performance evaluation, and model interpretation using SHAP. The experimental results showed that the proposed model achieved an accuracy of 68.42%, a precision of 67.79%, a recall of 68.42%, and an F1-score of 67.29%. Of the total reviews analyzed, 341 reviews (51.43%) were classified as positive, 208 reviews (31.37%) as negative, and 114 reviews (17.19%) as neutral. SHAP analysis revealed that the terms "banget," "bagus," "bikin," "Skintific," and "kulit" were the most influential features affecting the model's predictions of customer satisfaction. These findings provide useful insights for businesses to better understand customer perceptions and support continuous improvements in product quality and service performance.