摘要
目的:探讨Logitboost和累积比数Logit模型这两种方法应用于判别分析的优缺点。方法:简要介绍Logitboost和累积比数Logit模型的原理,并采用此两种方法分别对同一个实例进行判别分析。结果:两种方法的判别正确率均较高。Logitboost判别效果高于累积比数Logit模型判别。讨论:在迭代轮数适当的情况下,Logitboost判别正确率更高,受迭代次数影响较大;而累积比数Logit模型的稳定性较强。在对事件进行判别时,可根据数据资料的具体特点选用判别方法,也可将两种方法结合应用,取其判别效果较好者。
Objective： To compare Logitboost with Cumulative odds logit model, and discuss their characteristics when they are used in Discriminant analysis. Methods： The ultimate principle of Logitboost and Cumulative odds logit model will be introduced in this paper, and we will use the two methods to solve the same problem. Results Logitboost＇s effect is better than Cumulative odds logit model. Conclusion. The effect of Logitboost would be better if a appropriate iteration is given, in other words, Logitboost is affected by iteration in large measure. But Cumulative odds logit model is stable. We should choose the better according the data.
出处
《数理医药学杂志》
2007年第5期592-594,共3页
Journal of Mathematical Medicine
