A HYBRID MACHINE LEARNING MODEL FOR SMART BACKORDER RECOVERY AND PROACTIVE PRODUCT BACKORDER PREDICTION IN SUPPLY CHAIN MANAGEMENT AND INVENTORY OPTIMIZATION

Abstract

Product backorders remain a major challenge in Supply Chain and Inventory Management (SC&IMS), leading to inventory shortages, delayed order fulfilment, increased operational costs, and reduced customer satisfaction. This study proposes SMART Backorder Recovery, a hybrid machine learning framework for proactive product backorder prediction that integrates K-Means clustering with Naïve Bayes (NB), Logistic Regression (LR), and Stochastic Gradient Descent (SGD). The proposed framework combines unsupervised clustering with supervised classification to improve feature representation and predictive performance for highly imbalanced inventory datasets. Random Forest-based feature selection is employed to identify informative variables, while the Synthetic Minority Over-sampling Technique (SMOTE) is applied to address class imbalance. Model performance is evaluated using accuracy, precision, recall, F1-score, sensitivity, and specificity. Experimental results demonstrate outstanding predictive performance, with K-Means + SGD achieving 99.9996% accuracy, followed by K-Means + LR (99.9980%) and K-Means + NB (99.4884%). The findings demonstrate that hybrid machine learning substantially improves proactive product backorder prediction and provides an intelligent decision-support framework for inventory planning, operational resilience, and data-driven supply chain management.