一种信息不完全确定的多准则分类决策方法

A Multi-criteria Classification Approach with Incomplete Certain Information

针对准则权系数信息不完全确定和准则值信息不完全且有训练集的多准则分类决策问题,提出一种基于证据推理的分类方法.该方法在对训练集分类的基础上,结合不完全确定的准则权系数信息等建立非线性规划模型;然后利用遗传算法和单纯形法联合求解优化模型,得出准则权系数和分类效用阈值等参数,进而求出每一方案的效用值;最后与分类的效用阈值进行比较,得到方案集的分类.应用实例说明了该方法的有效性和可行性.

For a kind of multi-criteria classification problems with reference set, in which the information on the criteria weights is incomplete certain and criteria values is incomplete, a method based on evidential reasoning is proposed. Using classification to the reference set and incomplete certain information on weights, a nonlinear programming model is developed. Then by using genetic algorithms and simplex method to solve the nonlinear programming models, the optimal criteria weights, utility thresholds of categories and other parameters are obtained. And utility of the alternatives is computed by using interplation. Classification is performed through the comparison of the utilities of the alternatives to the utility thresholds of categories. Finally, an example shows the feasibility and availability of the method.