PREDICTIVE ANALYTICS FOR HIGH-VALUE CUSTOMER SEGMENTATION IN SUPPLY CHAIN MANAGEMENT

Abstract

Supply Chain Management (SCM) has become increasingly data-driven through machine learning (ML) techniques. The ability to precisely identify high-value customers requires accurate prediction methods which support strategic planning and marketing activities and operational efficiency. The research employs a complete machine learning process to analyze a specially designed dataset which includes both transaction data and customer information to discover high-value customers. We implemented and compared 11 machine learning models including Logistic Regression Decision Tree Random Forest Gradient Boosting SVM KNN Naive Bayes XGBoost LightGBM AdaBoost and Bagging Classifier. The performance evaluation used accuracy and confusion matrix and ROC-AUC score as assessment metrics for the models. The models achieved near-perfect performance with most of them reaching 100% accuracy and AUC scores. SHAP (SHapley Additive exPlanations) provided a method to explain model outputs while showing which features had the greatest impact. The research investigated unsupervised clustering and dimensionality reduction methods to achieve a better understanding of customer segmentation. The results demonstrate that ensemble-based models consistently achieved better performance than other models while SHAP analysis created better trust in model outcomes. The research findings present a useful framework which enables SCM predictive modeling to accurately identify essential customer segments.