ew 18: e14

Research Article

Improvised_XgBoost Machine learning Algorithm for Customer Churn Prediction

Download63 downloads
  • @ARTICLE{10.4108/eai.13-7-2018.164854,
        author={Swetha P and Dayananda R B},
        title={Improvised_XgBoost Machine learning Algorithm for Customer Churn Prediction},
        journal={EAI Endorsed Transactions on Energy Web: Online First},
        volume={},
        number={},
        publisher={EAI},
        journal_a={EW},
        year={2020},
        month={6},
        keywords={customer churn prediction, improvised-XG boost, telecommunication, prediction model},
        doi={10.4108/eai.13-7-2018.164854}
    }
    
  • Swetha P
    Dayananda R B
    Year: 2020
    Improvised_XgBoost Machine learning Algorithm for Customer Churn Prediction
    EW
    EAI
    DOI: 10.4108/eai.13-7-2018.164854
Swetha P1,*, Dayananda R B2
  • 1: Research Scholar, Visvesvaraya Technological University, Assistant Professor, Department of CSE, KS School of Engineering and Management, Bangalore, India
  • 2: Professor, Department of CSE, KS Institute of Technology, Bangalore, India
*Contact email: Shwetha6600@gmail.com

Abstract

The Customer retention has become one of the major issues for the service-based company such as telecom industry; where predictive model to observe customer, behavior is one of the efficient methods in the customer retention process. In this research work, ImprovisedXGBoost churn prediction model with feature functions is proposed, the main aim of this model is to predict the customer churn rate. ImprovisedXGBoost algorithm is a feature-based machine learning classifier which can be used for the complex dataset. At first, feature function is introduced then loss function is formulated and minimized through iterative approach, later combined with XGBoost approach it possesses better efficiency. The main feature of ImprovisedXGBoost algorithm is that it handles the unstructured dataset attributes efficiently, further feature function combined with XG_Boost. Furthermore, the proposed model is evaluated through various performance metrics such as accuracy, precision and recall. Our model also throws light on identifying the correctly and incorrectly classified instances on South Asia GSM (Global System for Mobile Communication) service provider. The results through the comparative analysis, our model outperforms the other state-of-art technique.