Modelo predictivo aplicado a la deserción de clientes de la corporación nacional de telecomunicaciones en la ciudad de Quevedo 2020 -2024

This study analyzed customer churn in the telecommunications services of the National Telecommunications Corporation (CNT) in the city of Quevedo, considering the period 2020–2024. The original dataset contained approximately 19,000 records, which, after cleaning, imputation, and quality validation processes, were consolidated into 18,484 observations useful for modeling. Data science techniques were applied following the CRISP-DM methodology, which included phases of business understanding, information exploration, preparation, modeling, and evaluation. The algorithms implemented included logistic regression, decision trees, Random Forest, XGBoost, and neural networks. The models were validated using metrics such as AUC, accuracy, sensitivity, and calibration curves, which allowed for the selection of optimal configurations to estimate the probability of customer churn. The results revealed factors associated with user churn, such as technical infrastructure, equipment failures, perception of rates, payment history, and service level. In addition, risk patterns were identified in specific areas, and profiles were constructed of customers most likely to cancel the service.

• Autore principale: Sánchez Carranza , Juan Carlos (author)
• Natura: masterThesis
• Lingua: spa
• Pubblicazione: 2025
• Soggetti: Deserción de clientes; Telecomunicaciones; Aprendizaje automático; Retención de usuarios
• Accesso online: https://repositorio.uteq.edu.ec/handle/43000/8806