Analisis Segmentasi Pelanggan dan Prediksi Churn pada E-commerce Menggunakan K-Means Clustering dan Random Forest: Studi Kasus Olist Brazilian E-commerce

Bayu Aditiya Nuryansyah (Universitas Pamulang)
Raihanullah (Universitas Pamulang)
Oriandika Rasyid (Universitas Pamulang)

Journal of Information Systems and Business Technology
Vol 2 No 3 (2026): Journal of Information Systems and Business Technology

The rapid growth of the e-commerce industry requires digital platforms to focus on customer retention strategies to ensure business sustainability. This study aims to integrate a customer intelligence approach through customer segmentation and loyalty risk prediction. The methods applied in this study combine unsupervised learning techniques using the K-Means algorithm and supervised learning using the Random Forest algorithm on the Olist Brazilian E-commerce dataset. The clustering process based on the Recency, Frequency, and Monetary metrics produced optimal groupings with a Silhouette Score of 0.36. Furthermore, the Random Forest model successfully predicted the potential for churn with an accuracy rate of 85.37%. The combination of these two methods significantly contributes to mapping high-risk customer segments, enabling management to formulate precise retention programs.