International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056 Volume: 11 Issue: 06 | Jun 2024
www.irjet.net
p-ISSN: 2395-0072
Customer Churn Prediction Analysis using Machine Learning Models Rishabh Kumar PG Student (Masters of Integrated Technology) Department of Computer Science and Engineering Noida Institute of Engineering and Technology Greater Noida Affiliated to AKTU Lucknow --------------------------------------------------------------------------***-----------------------------------------------------------------------
Abstract - Customer churn prediction is a critical aspect for
• Churn prediction decreases with tenure , people who spend more time with the company are likely to churn less.
businesses aiming to retain their client base in competitive markets. It is easier to retain a customer than to convert a new customer successfully This study applies advanced machine learning techniques to predict customer churn, leveraging a rich dataset with features including demographic information, service usage patterns, and customer account information. The analysis achieves robust prediction accuracy by employing methods such as SMOTE-Tomek resampling to address the class imbalance, and utilizing algorithms like LightGBM, Random Forest, Xgboost, etc. Feature importance techniques, including permutation importance and SHAP values, are employed to identify key factors influencing churn. This comprehensive approach provides actionable insights for targeted retention strategies, ultimately aiming to reduce churn rates and enhance customer satisfaction.
• People who use fiber optics are likely to churn mostly. • People having month-to-month contract prefer paying by Electronic Check mostly or mailed check. The reason might be short subscription cancellation process compared to automatic payment. • People with no internet service are least likely to churn followed by people who have online security and at last comes people who have highest chances of churning are those people who don’t have any online security despite having internet services:
Key Words: SMOTE-Tomek, Xgboost, LightGBM, Random forest, SHAP values, Hyperparameter tuning.
1. INTRODUCTION The globalization and advancements in the telecommunication industry have significantly increased market competition by introducing numerous operators. To maximize profits, companies adopt strategies such as acquiring new customers, up-selling, and extending the retention period of existing customers. Among these, retaining current customers is the most cost-effective.
Fig1. System Architecture
1.2 Machine learning models Random forest classifier:
The primary goal of customer churn prediction is to develop strategies for customer retention. With growing market competition, the risk of churn rises, making it essential to track loyal customers. Churn prediction models aim to identify early signs of churn and forecast customers likely to leave, helping companies leverage their valuable databases to maintain customer loyalty and mitigate churn.
• Random Forest is an ensemble learning method that builds multiple decision trees and merges them to get a more accurate and stable prediction. In churn prediction, it works by creating a large number of decision trees during training and outputting the class that is the mode of the classes of the individual trees. • Its robustness to overfitting and ability to handle highdimensional data make Random Forest a reliable algorithm for identifying churn patterns and predicting customer behavior.
1.1 Exploratory Data Analysis. • Performed EDA on churn datset to find out relationships bw various features that exist and conducted various statistical tests like t-test , anova test etc to confirm the relationships.
© 2024, IRJET
|
Impact Factor value: 8.226
|
ISO 9001:2008 Certified Journal
|
Page 222