粗糙集理论和遗传算法集成的上市公司违规行为预警研究  

The Early Warning of Listed Company's Illegal Behaviors Based on the Integration of Rough Set and Genetic Algorithm

作　　者：薛锋[1] 柯孔林[2] 

机构地区：[1]西安交通大学管理学院,西安710049 [2]浙江工商大学金融学院,杭州310018

出　　处：《软科学》2008年第4期49-53,共5页Soft Science

基　　金：国家自然科学基金项目(70171005)

摘　　要：以2002～2004年间受到监管部门处罚的上市公司为研究对象，经检验样本数据不服从正态分布及协方差矩阵不相等，故将粗糙集理论与遗传算法相结合，利用粗糙集方法具有不需要满足统计假设和遗传算法可以有效简化判别规则的特点，构建了符合上市公司数据实际情况的违规行为预警模型。实证结果显示该集成模型对上市公司发生违规行为的预测精度为82．00％，预警效果较好。Taking listed company punished by supervision department between 2002 - 2004 as sample, and considering the non - normal distribution and inequality of matrix of covariance feature, this paper combines rough sets with genetic algorithm. Rough set does not require statistics hypothesis and genetic algorithm may reduce discriminated rule, so this paper constructs a warning model about irregular activities suitable for listed company＇s data. It shows that integration model is an efficient and practical tool to forecast irregular activities of listed company.

关 键 词：上市公司 粗糙集理论 遗传算法 违规行为预警 

分 类 号：F275.5[经济管理—企业管理]