<![CDATA[PT Lintas Jaringan Nusantara, an internet service provider, faces challenges in utilizing customer data, which is mainly used for administrative purposes such as billing and support, limiting deeper analysis. This study applies K-Means clustering under the CRISP-DM framework for customer segmentation-based service-oriented attributes: internet package, price, and NAS location, using 972 customer records. Categorical attributes were transformed using frequency encoding and manual mapping. Model evaluation using the Elbow Method suggested 3 clusters, while the Silhouette Coefficient indicated that 10 clusters were optimal, improving the score from 0.5471 to 0.7704. The resulting clusters show variations in customer characteristics and provide an exploratory overview of grouping patterns. However, the 10 clusters solution should not yet operationally validated, as stakeholder validation involving marketing and customer service teams is still required to assess interpretability, business relevance, and practical applicability. Further validation using additional customer data or alternative datasets is also recommended. Overall, the findings serve as an initial analytical step to support future data-driven decision-making.]]>
Copyrights © 2026
