Data-Driven purchasing strategies: Price prediction models and strategy development

Abstract

This study aims to provide price predictions for automotive steel materials using machine learning (ML) methods to develop purchasing strategies with business theories. The primary objective is to respond to the understanding of purchasing dynamics, emphasizing the importance of adopting an agile approach to support the development of purchasing strategies within organizations. The data set used in the study was provided by a global original equipment manufacturer (OEM) company. Clustering analysis is performed to determine the most strategic project cluster, and price prediction models are developed for the most strategic project cluster using ML methods. Artificial neural networks (ANN), and tree-based models (decision trees (DT), bagging, and boosting methods) are used for price prediction models. According to the model's results, strategic purchasing suggestions are enhanced by incorporating a dynamic capability view (DCV) and information processing theory (IPT) to ensure adaptability and competitiveness in ever-changing purchasing dynamics. It was found that ANN performed the best despite its black-box nature. While tree-based models did not perform as well as ANN, they provided valuable insight into the importance of different criteria weights in price prediction. Integrating advanced ML techniques like ANN and tree-based models significantly improved price prediction accuracy. ANN parameters were carefully optimized, and decision tree structures allowed for make quick price predictions. Additionally, incorporating business theories such as DCV and IPT. The research enhances strategic purchasing recommendations, ensuring adaptability and competitiveness amidst evolving purchasing dynamics. These findings contribute to streamlining purchasing processes and emphasize the transformative potential of integrating business theories with ML methodologies in refining real-world analyses with precision.