一种属性约简方法及其在动员联盟伙伴选择中的应用

粗糙集的知识约简功能是针对离散数据的,而决策表属性取值很多是连续的.为此综合聚类与粗糙集方法,提出一种决策表的全局连续属性离散化、属性约简与规则提取的方法.动员联盟是敏捷动员的重要实现方式,动员联盟的伙伴选择指标体系体现了国家对动员企业的基本要求.在已有的动员企业潜力指标数据和经验决策数据的条件下,通过这种全局的属性离散化和属性约简方法,可以进一步精简指标体系,并提取有用的决策规则,经实验结果验证,使用全局属性约简方法处理动员联盟伙伴选择的指标约简具有更好的优越性.

A Global Attribute Reduction Algorithm and Its Application on Partner Selection of Mobilization Alliances

The knowledge reduction function of rough sets theory is specific on discrete data, while most decision tables are comprised of continuous attributes. Therefore a global discretization and attribute reduction algorithm based on clustering and rough sets theory was proposed. Mobilization alliance is important organization of realizing agile mobilization, the basic requirements of virtual enterprises are represented in the index system of partner selection. Under the condition of knowing the potential data of enterprises and decision results from case database, the index system can be reduced effectively through the global discretization and attribute reduction algorithm. An example was proposed to reduce the unnecessary indexes and induce the decision rules of partner selection of mobilization alliances. The results illustrate the feasibility and effectiveness of the algorithm.