An Extra Tree Classifier for prediction of End to End Customer Churn and Retention

• Published in: 2024 Asia Pacific Conference on Innovation in Technology (APCIT) pp. 1 - 6
• Main Authors: Gayathri, G., Priyanka, P, Manjeera, B, Niharika, K, S, Venkatrama Phani Kumar, K, Venkata Krishna Kishore
• Format: Conference Proceeding
• Language: English
• Published: IEEE 26.07.2024
• Subjects: Accuracy; AdaBoost; Adaptation models; Analytical models; Bagging with KNN; Customer Retention; Data models; Extra Tree Classifier; F1-score; Feature extraction; Nearest neighbor methods; PCA; Precision; Predictive Modeling; Predictive models; Recall; Ridge Classifier; XGBoost
• DOI: 10.1109/APCIT62007.2024.10673618

Customer retention remains a significant challenge in service-oriented sectors, critically impacting revenue and growth due to churn rates. This paper presents a comprehensive approach using multiple machine learning models to predict customer churn, leveraging a detailed dataset reflecting customer behavior and engagement. The methodology utilized includes diverse models such as bagging with K-Nearest Neighbors (KNN), Ridge Classifier, XGBoost, AdaBoost, and ExtraTree Classifier, with Principal Component Analysis (PCA) for feature extraction to enhance model input. Among these, the ExtraTree Classifier emerged as the most effective, achieving the highest accuracy and consistency across essential metrics like precision, recall, and F1-score. This indicates that the ExtraTree Classifier, paired with PCA, proficiently pinpoints the critical factors influencing customer churn. The findings provide actionable insights for businesses to formulate targeted customer retention strategies. Discussions highlight that the strategic application of the Ex-traTree Classifier, supported by robust feature extraction, can significantly influence customer management and retention initiatives.