不确定环境下基于数据挖掘的群体偏好行为评估

As the time of big data is coming, the behavior data of group preferences are emerging with the features of multi-source heterogeneous, low density of value as well as real-time changes. Because of uncertain environment and the limitations of experts’ knowledge, it is difficult for decision makers to integrate and analyze the big data. Therefore, based on the data mining techniques and the analysis of individual preference behavior of decision makers, a method for evaluating the behavior of group preference under uncertain environment is proposed in the project. Specifically, there are lots of breakthroughs: (1) Based on data mining techniques, the association rules between the heterogeneous data are studied in order to extract information for dynamic group decision making. (2) Analysis of the views and preferences of different individuals in group, and construction of the dynamic aggregation model for transforming from dynamic individual preferences to group preference. (3) According to different types of preference behavior data, an individual consistency test model and a consensus model for group preference behaviors are proposed based on different forms of information in hybrid-heterogeneous multi-attribute group decision making problems. This project will provide a theoretical basis for the application of big data to assist in the development of the group decision-making under uncertain environment. Moreover, the project conforms the reality of social development needs, but also makes for the rich and perfect decision-making theory and method system.

随着大数据时代的到来，多源异构、低价值密度、实时变化等特征的群体偏好行为数据不断涌现，加之环境的不确定性和决策者知识水平的有限性，决策者很难对多样性数据进行融合和分析。因此，本项目基于数据挖掘技术，通过对决策者的个体偏好行为进行分析，拟构建不确定环境下群体偏好行为的评估方法。在如下关键问题上取得突破：（1）基于数据挖掘技术，研究不确定环境下多源异构数据间的关联规则，为动态群体决策提炼信息；（2）分析群体中不同个体的观点和偏好，建立不确定环境下由动态个体偏好向群体偏好转化的动态集结模型；（3）针对不同类型的偏好行为数据，建立混合异构多属性群体决策中个体一致性检验模型和群体偏好行为的共识模型。本项目将为如何应用大数据，辅助开展不确定环境下的群体决策提供理论依据，既符合当今社会发展的现实需求，也有利于决策理论和方法体系的丰富和完善。

在决策过程中，决策的合理性和准确性直接取决于对不同环境下各个方案综合评价的合理性和科学性。如何在不确定的环境下，根据评估数据，分析专家的偏好行为，从而进行快速、高效的评估决策是一个重要的研究话题。本项目基于数据挖掘技术、模糊集理论和数学规划模型，对决策者个体偏好行为进行分析，构建了不确定环境下的群体偏好行为评估及大规模群体决策共识方法。本项目的主要研究成果体现在以下三方面：（1）针对不同类型的偏好行为数据，建立了混合异构多属性群体决策中的偏好集结模型、共识模型和反馈机制；（2）考虑不确定环境下决策者的特点，以决策者偏好信息的犹豫性为主线，结合模糊理论、多属性决策方法以及群体决策方法，构建了复杂多属性大群体决策模型；（3）结合多属性决策方法和数据挖掘方法，在准则层中综合考虑多个评价指标来评判基分类器的重要性，以实现对基分类器的评价和选择。本项目的研究成果为不确定环境下决策过程中的数据处理、偏好信息提取识别和群体共识提供了理论依据，丰富和完善了数据科学和决策科学的理论和方法体系。在项目执行期间，共发表学术论文9篇，其中SCI检索论文7篇，SSCI 检索论文1篇，中文核心期刊论文1篇，2篇论文入选ESI高被引论文，1篇论文入选ESI热点论文。