A multi-phase analytics framework for supply chain supplier selection and order allocation with delay risks and Industry 4.0 readiness

Numerous studies have addressed the Supplier Selection and Order Allocation (SSOA) problem, focusing on optimal quantity allocation. However, in practice, suppliers often fail to deliver allocated quantities on time due to operational delays or disruptions. Thus, incorporating supplier delays into order allocation decisions is essential. This paper introduces a multi-phase optimization framework that integrates the impact of delays into the SSOA process. In the initial phase, several Machine Learning (ML) algorithms are employed to predict delay probabilities at the order level. This study is the first to utilize ML-based delay probability predictions - rather than binary classification (on-time vs. delayed) - to determine optimal supplier allocations. The algorithms are evaluated using performance metrics such as accuracy, F1 score, precision, recall, and AUC, with TOPSIS used to select the most effective algorithm. Predicted probabilities are then aggregated to the supplier level for integration into the optimization model. Given the growing importance of Industry 4.0, the framework incorporates an Industry 4.0 Readiness Index (IRI), constructed using linguistic terms and interval numbers to handle subjective evaluations. The SWARA method is used to assign weights to evaluation criteria. These elements are embedded in a bi-objective optimization model, solved via the augmented ε-constraint method, aiming to minimize supply chain costs while maximizing suppliers' IRI scores. A numerical example based on a real-world case study validates the approach. Results show significant changes in supplier allocations when delay probabilities are considered, with a 4.84 % increase in total supply chain cost, primarily due to increased procurement in certain periods.