• 论文 •    

一类二次偏好聚合函数的多准则分类决策方法研究

李  蔷,高  阳,王坚强   

  1. 中南大学 商学院,湖南  长沙  410083
  • 出版日期:2005-12-15 发布日期:2005-12-25

Research on hierarchical discrimination approach for multi-criteria

LI Qiang, GAO Yang, WANG Jian-qiang   

  1. Sch. Of Business, Central South Univ., Changsha  410083, China
  • Online:2005-12-15 Published:2005-12-25

摘要: 求解多准则分类决策问题,虽有很多方法,但均存在一些不足。为此,提出了一种准则权系数不完全信息下的非线性多准则层次分类方法,该方法以层层递进方式对非训练集逐步完成整个分类决策过程。对于每一分类,首先通过训练集和偏好关系,建立一类二次偏好聚合函数的多准则分类优化模型;再利用基于进化策略和退火罚函数的求解算法求解优化模型,由此得到偏好聚合函数的系数等参数,计算每一方案的属于此类和不属于此类的一致性指标,进而确定方案是否属于此类。最后用实例验证了方法的可行性和有效性,得到了更贴近现实情况和科学合理的分类结果。

关键词: 多准则决策, 分类, 二次偏好聚合函数, 不完全信息

Abstract: To improve the existing methods for multi-criteria classification decision-making problems, a nonlinear hierarchical discrimination approach was presented for multi-criteria classification decision-making problems under correlated criteria and incomplete information. In this method, the whole decision process for classifying the alternatives to corresponding group was accomplished by the form of hierarchy. For each class, the first step was to build the multi-criteria classification optimization model through reference set and preference relationship. Then the optimization model would be achieved by algorithms based on evolutionary strategy and annealing penalty function. After that the coefficients of preference aggregation function and other parameters could be obtained, and the consistence index computation was conducted so as to discriminate and confirm the right class for each alternative. At last, scientific and reasonable results of a practical example have demonstrated that this approach was feasible, effective and close to the real situation.

Key words: multi-criteria decision making, classification, quadratic preference aggregation function, incomplete information

中图分类号: