考虑产品间协同效应的多产品组合采购问题研究

The multi-product combinatorial procurement problem considering the synergy between products is characterized by high complexity, inherent uncertainty, and difficult systematic. It has attracted considerable attention of the researchers in both academia and industry. The objective of this research is to group the required products into different synergetic product bundles consistent with buyer’s preferences and select candidate suppliers and cooperative suppliers for the product bundles, by theories and methods of ontology, cluster analysis, multi-criteria decision making, combinatorial procurement auction, multi-agent system, and so on. The research mainly covers the following topics: 1) establishing a ontology and rule based reasoning method to determine the preferences of the buyer; 2) developing a cluster-based approach to generate product bundles consistent with buyer’s preferences and supplies’ capabilities; 3) establishing a DEA-based supplier selection model to select candidate suppliers for product bundles; 4) establishing a multi-round combinatorial procurement auction model to determine final cooperative suppliers for product bundles, by considering the buyer and suppliers’ preferences on the synergy between products and supporting the interaction between the buyer and suppliers; 5) developing a multi-agent system to simulate the multi-product combinatorial procurement model. This research provides a promising methodology to solve the multi-product combinatorial procurement problem in consideration of the synergy between products, and hence offers great academic and practical potential in real-world procurement problem.

考虑产品间协同效应的多产品组合采购问题具有复杂性、不确定性、难以系统化等特点，是采购领域的重点和难点问题。本项目旨在运用本体、聚类分析、多准则决策、采购组合拍卖和多智能体等学科的相关理论和方法，把采购产品按采购商的偏好聚集成不同的协同性产品组合，基于这些产品组合选择候选供应商和确定最终合作供应商。主要研究内容是：基于本体和规则的情境推理法，建立采购商偏好确定模型；基于聚集型层次聚类法，建立符合采购商偏好和满足供应商供货能力的协同性产品组合确定模型；基于数据包络分析法，建立面向产品组合的候选供应商选择模型，为每种产品组合选择合格的候选供应商；基于采购组合拍卖方法，建立能够让采购商和供应商都表达产品间协同效应和支持采购商和供应商交互的多轮采购组合拍卖模型；基于多Agent的仿真方法，实现考虑产品间协同效应的多产品组合拍卖模型。研究成果将为解决多产品环境下的组合采购问题提供创新理论和方法。

本项目综合运用本体、聚类分析、多准则决策、采购组合拍卖和多智能体等学科的相关理论和方法，解决了考虑产品间协同效应的多产品组合采购问题。首先，通过提出产品间协同程度衡量指标体系和基于余弦相似度的产品间协同程度衡量方法，提出了协同性产品组合确定模型，该模型能够将采购产品聚类成满足采购商偏好的协同性产品组合。其次，通过提出面向产品组合的供应商评价指标体系和基于数据包络分析法的供应商评价方法，提出了面向产品组合的供应商选择模型，该模型能够为产品组合选择满足要求的候选供应商。此外，为解决数据包络分析法对供应商数目的要求，本项目进一步提出了基于熵权法和逼近理想解的排序方法的供应商评价方法。第三，通过提出基于简单加权法的投标评估函数、基于协同性产品组合的多轮采购拍卖机制、获胜者问题模型和基于分支定界法的获胜者问题求解算法，提出了基于协同性产品组合的多轮采购拍卖模型，该模型支持采购商和供应商对产品组合的多轮协商，进而为采购商选出最优合作供应商。最后，仿真实现了支持考虑产品间协同效应的多产品组合采购模型的多智能体系统。通过对比考虑和不考虑产品间协同效应两种情况下的实验结果，发现考虑产品间协同效应能够让采购商和供应商更充分表达其对产品间协同效应的偏好，能促进供应商提交更有竞争力的投标，进而能够使采购商获得更大的效用。因此，本项目提出的考虑产品间协同效应的多产品组合采购模型既是可行的也是有效的。