An Efficient Approach for Customer Attrition Prediction Using Regression-Based Feature Reduction and Neutrosophic Soft Set in Big Data Ecosystems
Abstract
The Neutrosophic Set (NS) concept is an overview of the theory of fuzzy sets and Indeterminant Fuzzy Sets (IFSs). NS is depicted by a truth Membership Function (MF), a false MF, and an indeterminant MF, and every membership level is a subset of the unit interval [−0, 1+]. Customers are the most significant strength of any organization and are reflected as the major source of income. Customer churn or attrition is a widespread phenomenon in several industries. Predicting customer attrition has great promise in improving customer retention. Recently, studies for customer churn prediction tend to employ Deep Learning (DL) approaches to handle huge amounts of data. This paper presents a Customer Attrition Prediction Using Regression-Based Feature Reduction and Neutrosophic Soft Sets (CAP-RFRNSS) model for big data ecosystems. The key objective was to explore the prediction of customer attrition to improve accuracy and scalability using an advanced neutrosophic model. In the data preprocessing stage, the standard scaler technique was employed to standardize the input. In addition, the Least Absolute Shrinkage and Selection Operator (LASSO) regression technique was deployed to select a feature subset to improve model generalization and reduce overfitting. Finally, for the customer attrition prediction process, the Possibility Interval-Valued Neutrosophic Soft Set (PIVNSS) method is implemented. A wide-ranging experimental analysis was conducted to determine the performance of the proposed CAP-RFRNSS system, demonstrating its supremacy with a maximum accuracy of 99.05%.