Enhancing Telecom Churn Prediction Using Emotion-Driven and Behavioral Engagement Features

Abstract: Accurate churn prediction enables service providers to develop effective retention strategies and promotes revenue stability in the telecommunication industry. This study enhances churn prediction performance by extracting five emotion-driven and behavioral engagement features from a telecom churn dataset. The new features represent derived, experience-oriented indicators constructed from operational usage data rather than direct psychological or survey-based measurements. To assess the effect of these engineered features on predictions, three powerful classifiers (i.e., CatBoost, Random Forest, and XGBoost) were trained and tested in a structured three-stage experimental design. In the first stage, the classifiers were trained and tested using the original dataset (original features only). In the second stage, the original dataset was enriched with five newly derived features (i.e., frustration index, trust score, satisfaction index, service usage score, and international experience index). Finally, in the third stage, only the engineered features were used in the classification process to evaluate their standalone predictive capability. Because the dataset is imbalanced, SMOTE and SMOTE-Tomek were applied to address this issue. The results demonstrate that incorporating these engineered features improves churn prediction performance across the reported evaluation metrics (accuracy, precision, recall, and specificity) for the classifiers and balancing techniques combinations presented. The enriched dataset (original + engineered features) achieves the strongest overall performance compared to using either original features only or engineered features only. Compared to the original features only, the enriched dataset achieved improvements of up to 3.6% in accuracy and 5.8% in recall. These findings indicate that emotion-driven and behavioral engagement features provide meaningful complementary information that enhances churn prediction effectiveness.

Huthaifa Aljawazneh. “Enhancing Telecom Churn Prediction Using Emotion-Driven and Behavioral Engagement Features”. International Journal of Advanced Computer Science and Applications (IJACSA) 17.2 (2026). http://dx.doi.org/10.14569/IJACSA.2026.0170284

@article{Aljawazneh2026,
title = {Enhancing Telecom Churn Prediction Using Emotion-Driven and Behavioral Engagement Features},
journal = {International Journal of Advanced Computer Science and Applications},
doi = {10.14569/IJACSA.2026.0170284},
url = {http://dx.doi.org/10.14569/IJACSA.2026.0170284},
year = {2026},
publisher = {The Science and Information Organization},
volume = {17},
number = {2},
author = {Huthaifa Aljawazneh}
}