Logistic模型选择中三种交叉验证策略的比较

The Comparison of Three Cross-Validation Strategy for Logistic Model Identification

在模型选择中,常用5折、10折交叉验证方法.文章给出一种基于3×2交叉验证的模型选择方法,并通过模拟实验证明了在Logistic模型中,3×2交叉验证要比5折和10折交叉验证选到真模型的概率更大.

The 5 fold and 10 fold cross validation methods are often employed in the model identification task.The 3×2 cross validation methods are introduced to identify the true model.The experimental results demonstrated that in the logistic regression framework,the probability of true model identification based on 3×2 cross validation is higher than that of 5 fold and 10 fold cross validation methods.