AN ADAPTIVE MACHINE LEARNING ARCHITECTURE OPTIMIZED FOR CUSTOMER CHURN PREDICTION

Abstract

The customer churn problem has a huge negative impact in different industries so that early prediction of customer churn can protect business by deciding early new business strategies that keep the customer before customers leave the service. The current research is enhancing the customer retaining by investigating architectural design model for predicting customer churn using Artificial Intelligence(AI). This research proposes adaptive Local Binary Social Spider Algorithm using Random Forest(LBSA-RF) model which is an architectural design model for predicting customer churn using artificial intelligence technique. The experiments took place on three datasets from different industries which are Orange dataset, IBM telecom dataset and bank churn dataset. The experiments administer the importance of real-time data analysis to capture a holistic view of customer interactions. The proposed architecture also concludes feedback for continuous model optimization and adaptation to changing customer patterns. The suggested Adaptive LBSA-RF model precision, recall, f1-score and accuracy performance metrics exceeds different related works when applied on the three mentioned datasets due to the intelligent automated feature selection mechanism of the Local Binary Social Spider Algorithm(LBSA) which effectively identifies and isolates the most predictive features, then leveraged by the Random Forest(RF) which is a powerful ensemble learning algorithm to build an accurate and robust classifier.