模型选择中交叉验证量的改进

The Refined Cross-validation Criterion for The Model Selection

对模型选择中交叉验证量CV进行改进,得到新的验证模型是否合适的准则RCV,RCV包含了CV的信息,并包含了拟合程度,模型中的待估参数个数和样本容量等等,比起AIC,BIC和CV具有更好的稳定性和分辨功能.

In this paper, the refined cross-validation value RCV for the model selection is given, it is better for the model selection than CV, AIC and BIC. RCV is the criterion that includes the residuals sum of squares, the number of sample and the number of parameters in the model. It is more stationary and distinguishable than CV, AIC and BIC.