基于行为挖掘和演化博弈的集中采购分组评标方法研究

For the realistic requirement of the bid evaluation in the centralized purchasing online of the government and enterprise, according to the multiattribute characteristics of objects, considering the influence of expert classifying and experts’ evaluating preferences to the fairness, this research focus on the bid evaluation experts’ selecting and the optimization of grouped bid evaluation decision-making by means of systems engineering, data mining and evolutionary game, aiming at improving the fairness of bidding evaluation. Firstly, we describe and quantify the behaviors and fairness of bidding evaluation, build the evaluation structure about fairness and cluster the behaviors of bidding evaluation based on the historical bid evaluation data. Secondly, we establish the evolutionary game model of bid evaluation and supervision behaviors based on behavior mining, analyze the evolutionary paths and factors influencing bid evaluation behaviors by adopting the concept of the evolutionary game. Lastly, we establish the optimization model of expert group’s formation according to the classification and evolutionary analysis. We propose the measure method of preference conflict and the mechanism of preference conflict’s dynamic resolution with exit-entrust-feedback. Subsequently, the model and decision support system of grouped bid evaluation decision-making based on choice preference are established. The study can not only make positive contributions to the enrichment and development of multiattribute group decision making theory, but can also provide theoretical basis and decision support for the government and enterprise about centralized purchasing, thus it has important practical significance to improve the fairness of bidding evaluation. The outcome can be also applied to other multiattribute decision making problems with similar grouped features.

面向政府和大型企业集团网络化集中采购评标的现实需求，针对物品的多属性特征，考虑专家分组及评标行为偏好对公平性的影响，采用系统工程思想、行为挖掘和演化博弈理论，重点研究评标专家的选取与分组评标决策方法的优化，以提高评标公平性。首先，描述和量化评标行为与评标公平性，建立公平性的评价结构框架，基于历史评标数据对专家评标行为进行聚类。其次，基于行为挖掘，建立评标与监管行为的演化博弈模型，分析行为演化规律及影响演化的因素。最后，依据行为聚类与演化，建立专家组组建的优化模型，构建专家群体偏好冲突的测度方法，建立“退出-委托-反馈”偏好冲突动态消融机制，建立基于选择偏好的分组评标决策模型与决策支持系统。该项研究不仅能丰富和发展多属性群决策优化理论，而且为政府和企业集中采购提供理论依据和决策支持，对提高评标公平性具有重要的现实意义。该成果还可应用于类似分组特征的其它多属性决策问题。

随着电子商务的飞速发展和“互联网+”行动计划的推进，网络化集中采购已成为政府和大型企业集团获取资源的重要途径，分组多属性评标是普遍采用的评标方法。本项目基于参与主体行为视角对集中采购中分组评标行为和评标决策方法等关键运作优化问题展开了系统性研究。项目完成了申请时所列出的全部研究内容，取得了一系列创新性成果，包括：考虑评标专家评标行为对集中采购评标公平性的影响，建立了评标专家与评标监管部门之间的博弈模型，基于有限理性特征，理论分析了演化均衡策略的存在条件及评标环境参数对均衡策略的影响，仿真分析关键参数及评标专家行为的影响机制；基于演化博弈理论，建立了高校科研工作者与管理部门的博弈模型，探索演化博弈方法论在行为运作管理中的应用，提出了促进评标行为向期望方向演化的管理措施；考虑网络效应和产品簇优化设计对集中采购决策的影响，基于离散选择模型建立产品动态定价优化模型，提出产品动态定价的优化策略；基于前景理论，建立了技术评标专家、商务评标专家和评标监管部门之间的三方博弈模型，理论分析专家群体的不同构成对评标结果的影响，提出促进评标专家向期望方向演化的优化策略；为支撑研究过程中的复杂模型求解，项目组还针对蝙蝠和鲸鱼等智能优化算法进行全局搜索能力的改进，有效提高最优解的全局性能。研究成果揭示了集中采购中分组多属性评标的客观规律，提出了提高评标公平性的优化策略和具体措施，为具有相关分组特性的多属性决策复杂问题的建模与优化提供了借鉴，同时可促进集中采购和招投标业务的健康可持续发展。