592 / 2024-04-25 23:06:42

Data-driven dual-sourcing multi-period procurement under price uncertainty

data-driven,procurement,price uncertainty,dual-sourcing

In the real world, it is essential for manufacturing firms to develop a raw material procurement plan over multiple planning periods so as to satisfy dynamic demands. However, they often face significant uncertainty in raw material prices. The prices of some raw materials may fluctuate greatly in a short time, and relying solely on the procurement in the spot market may expose firms to greater losses. Consequently, to reduce the losses caused by price fluctuations, we investigate a dual-sourcing strategy involving both spot and future procurement. Prices in the future procurement are always uncertain, but with abundant relative data. To address these issues, the predict-then-optimize (PTO) framework, a classical two-stage model, is commonly used. However, the separation between prediction and optimization may result in the loss of valuable information and lead to prediction errors, significantly impacting decision-making. Therefore, our study provides a data-driven framework for the multi-period procurement problem with both spot and future procurement under price uncertainty, where a price prediction model is proposed with the aims of minimizing decision loss. Through conducing the numerical experiments with real data, we demonstrate that our framework effectively reduces costs compared to those incurred by single-source procurement and the PTO model. Furthermore, our framework is able to extended to more complex procurement situations.