An Intelligent Predictive Model for Customer Experience Evaluation in Telecom Networks

Abstract

The rapid expansion of digital services in the telecommunications industry has significantly increased customer expectations regarding network quality, reliability, and service responsiveness. Customer experience has become a critical factor influencing customer loyalty and business sustainability. However, conventional customer satisfaction evaluation methods are mostly reactive and rely on descriptive statistics, which limits their ability to support proactive service improvement. This paper proposes an intelligent predictive model for evaluating customer experience in telecom networks using machine learning techniques. The proposed framework integrates network Quality of Service (QoS) parameters, including throughput, latency, jitter, packet loss, and service downtime, with customer-related variables such as complaint frequency, response time, and service tenure. Two supervised learning algorithms, Random Forest and Naïve Bayes, are employed to classify customer experience into satisfied and unsatisfied categories. The dataset is processed through data cleaning, normalization, and feature encoding, followed by model training and testing using a 70:30 data split and k-fold cross-validation. The experimental results demonstrate that machine learning models can predict customer experience with high accuracy and stability. Random Forest achieves superior performance and provides feature importance analysis, while Naïve Bayes offers efficient probabilistic classification. The results confirm that both network performance indicators and customer interaction data significantly influence customer experience. This research provides a data-driven framework that enables telecom operators to improve service quality and customer retention strategies proactively.