Potential Wi-Fi Customer Segmentation System Based on K-Means and WebGIS

Identifying potential IndiBiz Wi-Fi customers at PT Telkom Indonesia Pemalang Area requires a structured, measurable approach, as customer data is distributed across multiple geographic locations and is difficult to prioritize manually. In the case study, the initial dataset comprised 5,070 Point of Interest (POI) records, indicating that manual verification would be inefficient if all prospective customers were given the same priority. This study proposes a web-based segmentation system that integrates K-Means clustering, Optical Distribution Point (ODP) association analysis, WebGIS visualization, and Scrum-based system development into a single operational platform. The system uses Point of Interest (POI) data, including longitude, latitude, and odp_count, as clustering features to represent spatial location and infrastructure associations. The initial dataset consisted of 5,070 records and 12 attributes. After preprocessing, 5,001 valid records were used for clustering. The evaluation results showed that the best segmentation was obtained using K=3, with a silhouette score of 0.4833 and a Davies-Bouldin Index of 0.8116. The clustering results produced 2,771 medium-potential records, 1,525 low-potential records, and 705 high-potential records. The high-potential segment accounts for 14.10% of valid customer candidates and can serve as an initial priority list for field verification.