@article{4669,
  author = {Dit Suthiwong},
  title = {A Dual-Method Framework for Churn Prediction and Customer Segmentation in Telecommunications Using SVM and K-Means Clustering},
  journal = {Signals and Telecommunication Journal},
  year = {2026},
  volume = {15},
  number = {1},
  doi = {https://doi.org/10.6025/stj/2026/15/1/29-39},
  url = {https://www.dline.info/stj/fulltext/v15n1/stjv15n1_3.pdf},
  abstract = {This study investigates machine learning approaches for customer churn prediction and segmentation in the
telecommunications sector, addressing the critical business challenge of subscriber retention amid rising
acquisition costs. Leveraging the Telco Customer Churn dataset, comprising 7,043 customer records with
21 demographic, service usage, and billing attributes, we implement a dual method framework that combines
supervised classification and unsupervised clustering. A Support Vector Machine (SVM) with a radial basis
function kernel models nonlinear relationships between customer attributes and churn behavior, achieving
79.18% accuracy and a ROC-AUC of 0.797. As an interpretable benchmark, Logistic Regression delivers
superior ranking performance (ROC-AUC: 0.84) with 80.77% accuracy and a churn recall of 0.578, enabling
actionable insights through coefficient analysis. Complementing prediction, K-means clustering identifies
three distinct customer segments High Value Loyalists, Budget Newcomers, and Mid Tier Growth based on
tenure, charges, and service adoption patterns. Cluster validity is confirmed through silhouette analysis
(optimal K=3, score 0.52) and PCA visualization. The segmentation reveals strategic opportunities: loyalty
programs for high value customers, onboarding support for at-risk newcomers, and cross selling for growthoriented
mid tier users. Results demonstrate that effective churn management requires integrating predictive
accuracy with nuanced customer understanding. While nonlinear models capture complex behavioral
patterns, interpretable linear models offer practical advantages for operational deployment. This research
provides a robust analytical foundation for data driven retention strategies, enabling telecom providers to
transition from reactive interventions to proactive, personalized customer relationship management that
enhances long term profitability in competitive markets.},
}