Proposal for a Customer Recommendation and Retention System for the Telecommunications Sector in Peru Using a Hybrid Bayesian and Knowledge-Based Agent

Hidalgo K.M.O. · Proceedings of the 2025 IEEE 32nd International Conference on Electronics, Electrical Engineering and Computing, INTERCON 2025 (2025) · DOI: 10.1109/intercon67304.2025.11244691

Resumen

This paper proposes a hybrid intelligent system for customer retention in the Peruvian telecommunications sector, combining a Bayesian classifier and a knowledge-based agent within an event-driven architecture. The system can identify users at high risk of churn in near real-time and generate personalized retention recommendations, achieving an average latency of 2 seconds and computational complexity of O(1) for prediction and O(r × c) for rule evaluation. The solution follows the CRISP-DM methodology, processing billing, usage, complaints, and top-up data through big data tools such as Kafka and Spark. It also offers explainability through an editable decision table designed for business experts. In tests using 5,000 simulated records, the classifier achieved an F1-score of 0.76 for the churn class, demonstrating high sensitivity despite class imbalance. Simulations confirm that the system optimizes incentive allocation, reduces churn, and enables fast decision-making without requiring major investments. This proposal presents a viable and scalable alternative for emerging environments with limited infrastructure, positioning itself as an effective explainable AI solution for real-time personalization and improved commercial sustainability for telecom operators.

Instituciones: Universidad Nacional de Ingeniería