Computational Experiment-Driven Surrogate-Assisted Multi-objective Decision-Making Method For Complex Supply Chains

Supply chain is a complex multi-evolutionary system,and its evolution process has path dependence,irreversibility,uncertainty,and bifurcation of evolutionary trend.These characteristics increase the difficulty of managers’ decision-making and challenge the effectiveness of decision-making,especially in the supply chain scenarios with rapid decision-making requirements.In related work to supply chain decisions,operations research,control theory and other traditional analytical method are difficult to modeling complex supply chain multivariate evolution process;data security and the relative scarcity limit the machine learning method to be applied in some areas;though simulation can predict the complex evolution of supply chain system and the corresponding simulation optimization method can extend decision-making from prediction to prescription,simulation and simulation optimization still faces the problem of insufficient exploration on the potential multivariate emergence and evolution in supply chains.Since computational experiments can explore multiple emergent phenomena through constructing "virtual laboratories",this methods show great potential in the field of complex supply chain decision-making,but the relevant research is still generally limited in predictive decision-making.To cope with the complex evolutionary multi-objective decision of supply chain and extend the decision-making from prediction to prescription,this study employs the computation experiment as the approximation of multi-objective optimization functions,focuses on the effectiveness of multivariate emergence and the stability behind the multivariate emergence,proposes a computation experiment-aided multi-objective decision-making method.Furthermore,for the supply chain scenarios with fast decision-making requirements,this study implements surrogate models trained by machine learning methods driven by computational experiments as the approximation of multi-objective optimization functions,and proposes a computational experimentdriven surrogate-assisted multi-objective decision-making method in complex supply chains to ensure the effectiveness of the decision-making and achieve decision efficiency.Firstly,owing to the deficiency of simulation optimization methods in multivariate emergence exploration,this study proposes a computational experiment-aided multiobjective decision-making method for complex supply chains.This method supports the virtual scenario reproduction of the real supply chain by constructing computation experiment models,and emerges the multivariate potential evolution of the supply chain via multi-scenario calculation.Based on the multivariate evolution results,the stability evaluation behind the diversified evolution results is innovatively introduced into the multiple objectives in the supply chains,and support the generation of Pareto optimal solution sets with the integration of multi-objective evolutionary algorithms(NSGA-Ⅱ,NSGA-Ⅲ,etc.).Considering the dimensional difference of multiple objectives and ensuring the objectivity of decision-making,the Entropy-TOPSIS method is employed to prioritize the Pareto optimal solution set for providing supply chain managers with alternative decision-making solutions.This study proposes a multi-generation smartphone marketing case of some company,based on the multivariate emergence of the company’s total revenue,innovative technology acceptance degree,and consumer total utility,optimizes and prioritizes the pricing,production,and advertising decisions.The case results demonstrate that the proposed method can achieve the optimal solution generation based on the stability improvement behind the complex supply chain multivariate emergence.Secondly,relying on the effectiveness of computational experiments in supporting decision-making,to further meet the demand for rapid decision-making in supply chain scenarios,this study progressively proposes a computational experiment-driven surrogate-assisted multi-objective decision-making method for complex supply chains.The method employs computational experiments to reproduce the supply chain multivariate emergence data,drive surrogate models(Support Vector Machine,Gaussian Process Function,etc.)for training,testing,and generating,integrate the capacity of computational experiments in exploring multivariate emergence and the efficient predictive capacity of surrogate models,and employs the surrogate model as the approximation of computational experiment.The Pareto optimal solution set is generated through the integration of multi-objective prediction based on surrogate models and solution optimization based on multi-objective evolutionary algorithms(NSGA-Ⅱ etc.).This study proposes a multi-generation drink marketing decision-making case of competitive companies,construct computational experiment based on social networks for multivariate emergence data generation,and validate the predictive effectiveness of computational experiment with the empirical Norton-Bass model.The constructed computational experiments drive the surrogate model generation(Support Vector Machine),and optimization of introductory discount and discount period strategies.Case results indicate that the proposed method can achieve the efficient generation of optimal strategies under the premise of effective multi-objective decision-making in complex supply chain.The main contribution of this study is threefold:(1)For the decision-making under complex supply chain multivariate emergence,this study innovatively proposes a computational experiment-aided multi-objective decision making method.This method employs the potential of computational experiments in exploring multivariate emergence,sufficiently exploits the solution stability behind multivariate emergence,enriches and develops the theory and method of complex supply chain simulation optimization.(2)This study progressively proposes a computational experiment-driven surrogate-assisted multi-objective decision-making method,which combines the effectiveness of computational experiment in exploring multivariate emergence and efficient prediction of surrogate models through the integration of computational experiments,machine learning,and evolutionary algorithms.The proposed method improves the quality and efficiency of multi-objective prescriptive decision-making in complex supply chains and enriches the research system of complex supply chain multi-objective decision.(3)The proposed method provides managers with a novel decision-making perspective and efficient decision-making tools,improves the decision-making in supply chains from the perspective of complex systems,and supports enterprises to implement the intelligent operation management with an agile response to market changes.