摘要
由于数据的复杂性,一般的投影判别方法在应用上有一定的局限性,本文引入权重因子对不同投影方法进行组合,从而提出一种组合判别方法,以提高判别回代正确率与预测效果.通过实际例子和模拟数据验证了组合判别模型的判别效果良好,回代正确率比单一判别方法高,预测效果也优于单一判别方法.
Because of the complexity of the data, the general projection discriminant method suffers from certain limitations in appli- cations.For the purpose of taking correct rates of back substitution as the standard, this article gives a combination discriminant method by introducing weighting factors and combining different projection in order to improve accuracy and prediction effects.Simu- lated data are checked,and it is found that their correct rates of back substitution are better than those of the general linear discrimi- nate method.The prediction effect is superior to other methods.
出处
《厦门大学学报(自然科学版)》
CAS
CSCD
北大核心
2016年第3期366-370,共5页
Journal of Xiamen University:Natural Science
基金
国家自然科学基金(11301084)
福建省自然科学基金(2014J01010)
关键词
FISHER判别
内聚外散
回代
正确率
权重
组合判别
fisher discriminant
assembling inside and spreading
back substitution
correct rates
weighting
combination discrimi- nant