# Customer churn prediction in telecom using machine learning and social network analysis in big data platform

**Authors:** Abdelrahim Kasem Ahmad, Assef Jafar, Kadan Aljoumaa  
**arXiv:** [1904.00690](https://arxiv.org/abs/1904.00690)  
**Date:** 2019-04-01  
**Summary page:** https://www.summarizepaper.com/en/arxiv-id/1904.00690v1/

## Key points

- Customer churn is a major concern for large companies in the telecom industry as it directly impacts revenues.
- Companies are increasingly using predictive analytics to identify at-risk customers to combat customer churn.
- The study focuses on developing a churn prediction model using machine learning techniques and social network analysis (SNA) on a big data platform.
- The key contribution of the research is the development of a predictive model that assists telecom operators in identifying customers most likely to churn, achieving an AUC value of 93.3%.
- Incorporating SNA features further enhances the model's performance, increasing the AUC from 84% to 93.3%.
- The study utilized a large dataset provided by SyriaTel telecom company spanning nine months for testing and validation of the model.
- Four different algorithms were experimented with, with Extreme Gradient Boosting (XGBOOST) emerging as the most effective classifier for churn prediction.
- Advanced analytics techniques can effectively address customer churn in the telecom sector by leveraging big data and machine learning.

## Summary

Customer churn is a major concern for large companies in the telecom industry as it directly impacts revenues. To combat this issue, companies are increasingly turning to predictive analytics to identify at-risk customers. This study focuses on developing a churn prediction model using machine learning techniques and social network analysis (SNA) on a big data platform. The key contribution of this research lies in the development of a predictive model that assists telecom operators in identifying customers most likely to churn. By leveraging machine learning algorithms and innovative feature engineering and selection methods, the model achieves an impressive Area Under Curve (AUC) value of 93.3%. Additionally, incorporating SNA features further enhances the model's performance, increasing the AUC from 84% to 93.3%. To test and validate the model, a large dataset provided by SyriaTel telecom company was utilized. This dataset contained comprehensive customer information spanning nine months and served as the basis for training, testing, and evaluating the predictive system. The study experimented with four different algorithms - Decision Tree, Random Forest, Gradient Boosted Machine Tree (GBM), and Extreme Gradient Boosting (XGBOOST) - with XGBOOST emerging as the most effective classifier for churn prediction. Overall, this research showcases how advanced analytics techniques can effectively address customer churn in the telecom sector by harnessing the power of big data and machine learning. Telecom companies can proactively identify at-risk customers and implement targeted retention strategies to mitigate revenue loss.