Abstract

This study proposes an integrated agile decision-support model for predicting stockout risks in the e-grocery industry using machine learning techniques. The research aims to enhance inventory responsiveness and adaptability—two central pillars of agile management—by transforming operational data into predictive intelligence. A publicly available dataset comprising 1,000 Stock Keeping Units (SKUs) with 37 variables was utilized, encompassing attributes related to inventory, supplier performance, and demand forecasting. After data cleaning, transformation, and feature engineering, a set of eleven operational features was used to train and compare three classification models: Logistic Regression, Random Forest, and XGBoost. The models were benchmarked through Accuracy, Precision, Recall, F1-score, and ROC–AUC metrics. Results show that ensemble-based methods, particularly XGBoost, achieved superior performance, attaining 100% accuracy, precision, recall, and AUC. SHAP (SHapley Additive exPlanations) analysis revealed that Stock_Risk_Index, Lead_Time_Days, Forecast_Next_30d, and Safety_Stock were the most influential variables, capturing both supply responsiveness and adaptive buffer mechanisms. These findings confirm that operational agility is quantifiable through data-driven modeling and that machine learning can replicate and enhance decision rules traditionally applied in inventory management. The model was subsequently embedded into an agile decision-support cycle consisting of prediction, decision, and feedback stages. This integration enables continuous learning, real-time visibility, and transparent decision-making—core components of agile management frameworks. While perfect accuracy indicates deterministic relationships rather than stochastic generalization, the framework remains an effective validation of agile intelligence principles. The study concludes that data-driven learning systems, when combined with explainable AI techniques, provide a scalable pathway for embedding agility, responsiveness, and transparency within supply chain decision-making processes.

Keywords: Agile Management, Machine Learning, Stockout Prediction, Supply Chain Analytics, Decision Support System

How to Cite:

Nurdiyanti, O. & Afuan, L., (2026) “Predicting Stockout Risks Using Machine Learning for Agile Decision Support in the E-Grocery Supply Chain”, Agile Management 1(3), 169-184. doi: https://doi.org/10.63913/am.v1i3.134