面向大数据的群体偏好决策分析研究

In the network information society, multi-source, heterogeneous, large emerging real-time big data record human beings' various behavior and mental activities as well as a variety of different groups' preferences etc. Since these preference information have the following complicated features such as diversity, real-time changes and veracity, it becomes an important research subject of group decision-making under the big data environment that how to mine the useful group preferences from the big data and make group decision making. This project will focus on group preferences in big data, and attempts to study the following issues, including the method of intelligently capturing the individual point of view, the models of efficiently dealing with and intelligently mining the dynamic veracity data of individual preferences, the interaction mechanism of individual perspectives and individual preferences, the method of dynamic assessment and prediction on group preferences, the models of group preference clustering and fast matching, group preferences based group decision-making etc. We mainly concentrate on group preferences, and will build new theorems and methodologies of group preferences decision analysis in big data by integrating opinion dynamics, data mining techniques as well as decision theories. The results of this research project may have important theoretical significance and practical value since it cannot only enrich the theories and methodologies of decision making, but also can help the government and enterprises make more scientific and finer decision, operation and management by taking full use of groups preferences.

在网络社会中，多源异构、实时涌现的大数据记录了人类的各种行为和心理活动等信息。由于这些信息具有多样性、实时变化和低价值密度等复杂特性，如何从这些数据中挖掘出有用的群体偏好信息并进行群决策是大数据环境下的重要研究课题。本项目以大数据下群体偏好为研究对象，重点研究个体观点的智能获取、个体偏好动态多样性数据的高效处理和智能挖掘、个体观点与个体偏好的相互作用机理、群体偏好动态评估与预测、群体偏好聚类和快速匹配、基于群体偏好的群决策方法等。本项目以群体偏好为主线，结合舆论动力学理论、数据挖掘技术和决策理论方法，理论研究与实证研究相结合，构建面向大数据的群体偏好决策分析新理论与新方法。此外，研究成果有望对政府和企事业单位结合群体偏好信息进行更加科学和精细化的管理、经营及决策，具有重要的实际应用价值。

互联网的快速发展带来了海量数据，利用大数据进行群体偏好的获取和挖掘能够带来更具有整体性、客观性和实时性的分析，从而带来更精准、高效和科学的决策。但大数据的多样性、实时性和价值密度低等特性为群体偏好的分析与挖掘带来了挑战。为了从大数据中挖掘有用的群体偏好信息并进行群决策，本项目着重针对数据挖掘与优化、群体偏好分析与整合以及群体决策行为的演化的基础理论，进行相关应用研究。在数据挖掘与优化方面，在市场结构可视化、预防维修政策可视化、金融市场风险预测等多个方面开发了优化与挖掘方法；在群体偏好分析与整合方面，实现了具有格序偏好、异构信息、乘法和模糊偏好关系等特点的群决策整合方法；在群体决策行为的演化方面，研究了有噪声的、具有极端意见的以及社交网络环境下的群决策演化过程。除此之外，开展了大数据环境下精准扶贫等方面的应用研究。在相关领域的国内外期刊发表论文21篇，其中SCI检索14篇，SSCI 2篇。