基于离差优化的信息不完全确定的多准则分类方法

提出一种准则权系数信息不完全确定且有训练集的多准则分类决策方法.该方法利用训练集的分类结果和准则权系数的不完全确定信息,基于类间离差最大和类离差最小构建非线性规划模型,再利用遗传算法求解所得优化模型,得到训练集中各方案在各准则下的效用值和准则权系数,通过插值得到方案集中方案在准则下的效用值,确定方案与各类间的离差,进而得到方案的分类.实例计算表明了该方法的有效性和可行性.

Multi-criteria Classification Approach with Incomplete Certain Information Based on Optimizing Deviation of Categories

A multi-criteria classification approach with incomplete certain information and the reference set base on optimizing deviation of categories is proposed. Using classification to reference set and incomplete certain information on weights, nonlinear programming is constructed based on maximizing deviations between categories and minimizing deviations on categories. The criteria weights and utilities of alternative in reference set under criteria are obtained by using genetic algorithms to solve the nonlinear programming model. The utility of alternatives is gained by using interpolation. The deviations of alternatives and categories are computed, and the alternatives are classified. An example shows the feasibility and availability of the proposed approach.