Deep Learning-Driven Dynamic Segmentation and Sentiment Prediction to Enhance Customer Retention in Online Platforms

Abstract

Customer retention and churn prediction are critical in business analytics, requiring advanced methodologies to understand and predict customer behavior. This study presents an integrative framework that combines sentiment analysis and customer segmentation to address these challenges. A sentiment prediction model, trained on Amazon customer reviews, classifies sentiments as Positive or Negative, providing insights into customer perspectives. A segmentation model then categorizes customers into five loyalty groups—Champions, Loyalists, Potential Loyalists, At-Risk, and Detractors—based on demographic and behavioral data. The framework dynamically integrates new customer data, updating loyalty labels through sentiment analysis to identify potential churners. By blending sentiment-driven insights with segmentation dynamics, this approach offers a scalable and adaptive methodology for churn prediction. This research contributes to the literature on customer analytics, providing a practical tool for enhancing customer retention strategies [1-3].