Implementation of Machine Learning in Business Intelligence for Customer Segmentation and Loyalty at PT. Inti Group

Abstract

This study addresses the need for integrated data analytics and machine learning in PT Inti Group’s BI dashboard by implementing an unsupervised K‑Means clustering method on historical training data (January 2021–May 2025) extracted directly from a PostgreSQL database and analyzed using Python. The analysis process includes data preprocessing and feature engineering to create key variables: number of participants, training‑type frequency, recency (days since the last training), and engagement duration. Cluster determination was evaluated using the Elbow method (4 clusters), Silhouette score (2 clusters), and Davies–Bouldin index (9 clusters). Based on business interpretation and the balance between cluster compactness and separation, four clusters were selected: Loyal & High‑Value Customers, Inactive, Growing/Potential, and New/Sporadic. Customers who attended training more than ten times were classified as loyal. The segmentation results are visualized in a Power BI dashboard integrated directly with the data source, supporting rapid data‑driven managerial decisions. This study demonstrates that integrating unsupervised learning with BI effectively enhances understanding of customer characteristics and serves as a basis for designing more targeted marketing strategies. A limitation of this study is that the data cover only up to May 2025.