Machine Learning Models for Predicting and Clustering Customer Churn Based on Boosting Algorithms and Gaussian Mixture Model

Customer churn prediction is becoming increasingly important in the telecom sector, as churn is a significant driver of EBITDA (Earnings Before Interest, Taxes, Depreciation, and Amortization) margin. Processing of vast consumer data has made churn prediction and factor identification for customer r...

Full description

Saved in:
Bibliographic Details
Published in2022 International Conference for Advancement in Technology (ICONAT) pp. 1 - 5
Main Authors Vakeel, Aditi, Vantari, Neha Reddy, Reddy, Sai Nivas, Muthyapu, Rishith, Chavan, Ameet
Format Conference Proceeding
LanguageEnglish
Published IEEE 21.01.2022
Subjects
Boosting
Clustering algorithms
CRM
Customer Churn
GMM
LightGBM
Machine Learning
Machine learning algorithms
Prediction algorithms
Predictive models
Real-time systems
Telecommunications
Online AccessGet full text

Cover

More Information
Summary:Customer churn prediction is becoming increasingly important in the telecom sector, as churn is a significant driver of EBITDA (Earnings Before Interest, Taxes, Depreciation, and Amortization) margin. Processing of vast consumer data has made churn prediction and factor identification for customer retention a challenging task. The proposed work presents an effective Machine Learning (ML) approach that includes boosting algorithms to identify Customer Churn (CC) and Gaussian Mixture Model to cluster the churned customers. The proposed work also implements the Light Gradient Boosting (LightGB) model, which outperforms Adaboost and XGBoost for churn prediction by a factor of 15x. Further, the Gaussian Mixture Model (GMM) opted for clustering yielded a silhouette score of 0.36.
DOI:10.1109/ICONAT53423.2022.9725957