A data-driven decision-support framework for supply management using multi-criteria evaluation and machine learning

This paper presents a data-driven framework for supply management based on curated operational data across the end-to-end supply process. Data is maintained through automated validation, cross-checking, and traceability, enabling reliable analysis. Building on this foundation, three analytical flows are established: expert-based evaluation using the Analytic Hierarchy Process (AHP), predictive modeling with machine learning, and exploratory unsupervised analysis. For each flow, task-specific datasets and feature representations are derived to address managerial objectives. The framework is validated using interconnected request, procurement, and receiving data. From these records, AHP evaluates and classifies suppliers using multiple criteria, while machine learning predicts resulting performance classes from operational features. Results show that enriching supplier-item-type representations with aggregate item-type information generally improves predictive performance, highlighting the value of contextual feature engineering. Exploratory clustering reveals patterns in employee workload and processing efficiency. Collectively, findings demonstrate how the framework transforms operational data into actionable insights for supply management.