Data-driven decision making for supply chain networks with agent-based computational experiment

摘要

The complicated micro structures, macro emergences and dynamic evolutions in a supply chain network pose challenges to decision making for solving operational problems for the network's performance improvement. Most of these problems are complicated since various factors and their complicated relationships are involved. Success of this decision making relies on efficient business analytics based on the comprehensive and multi-dimensional data related to the static attributes and dynamic operations of the network. To confront the challenges, this paper proposes to explore a methodology of data-driven decision making for supply chain networks. In this methodology, a data-granularity model of a supply chain network is developed to standardize the data form for decision making. A four-dimensional-flow model of a supply chain network is proposed to satisfy the data requirements for decision making that are defined in the data-granularity model. Agent-based computational experiment is employed to support the generation of a comprehensive operational dataset of a supply chain network and to verify the solutions generated in decision making. Integrating these models, a data-driven decision-making framework for supply chain networks is proposed. In the framework, a new decision-making mode of "problem definition - business analytics-solution verification-parameter adjustment" is proposed. Oriented towards domain knowledge in supply chain networks, two approaches of business analytics mapping analysis and correlation analysis are presented. Finally, a case of a five-echelon manufacturing supply chain network is studied with the methodology. The findings indicate that the proposed methodology, models and framework are effective in supporting the data-centric decision making for solving complicated operational problems in supply chain networks and provide the networks' managers or member enterprises with an effective tool to generate unbiased and efficient decisions for the networks' performance improvement. (C) 2017 Elsevier B.V. All rights reserved.