基于EM的不平衡数据关键质量特性识别  

Identification of Critical-To-Quality Characteristics in Imbalanced Data Sets Based on EM Algorithm

作　　者：闫伟[1] 何桢[1] 田文萌[1] 何曙光[1] 

机构地区：[1]天津大学管理与经济学部,天津300072

出　　处：《工业工程与管理》2012年第4期38-42,共5页Industrial Engineering and Management

基　　金：国家自然科学基金重点项目(70931004);国家科学自然基金资助项目(71002105)

摘　　要：为了在高维不平衡质量特性数据集中提高关键质量特征识别效率,将EM(ExpectationMaximization)算法引入,通过逐步缩小比例较大数据集内的样本数量来抵消数据不平衡带来的负面影响。算例表明,该方法具有一定的合理性和可行性。In order to improve the efficiency of identifying the critical quality characteristic in data sets with high dimensional and imbalance,the Expectation Maximization Algorithm was introduced to offset the negative impact from the imbalanced data by reducing number of samples in larger class data set.The experimental results show that the method is sensible and feasible.

关 键 词：复杂产品 聚类 高维 不平衡数据 

分 类 号：F406.3[经济管理—产业经济]