陕西科技大学学报（自然科学版）Issue(1)：179-183,5.
基于组块3×2交叉验证 t检验的模型选择算法
Model selection algorithm based on blocked 3 × 2 cross-validated t-test
摘要
Abstract
In traditional model selection problem ,the evaluation of the model is always based on the performance indexes .However ,in many cases ,the difference of the performance inde‐xes of the models may not be statistically significant ,which may result from random error . Therefore ,a new model selection algorithm is proposed based on blocked 3 × 2 cross‐validated significance t‐test for the difference of the indexes .Experiment results show that the algo‐rithm always select the simple models based on the minimization of square loss (regression) or 0‐1 loss (classification) .关键词
模型选择/组块3×2交叉验证/t检验/测试误差/模型复杂度Key words
model selection/blocked 3 × 2 cross-validation/t-test/test error/model com-plexity分类
信息技术与安全科学引用本文复制引用
刘焱青,王钰,李济洪..基于组块3×2交叉验证 t检验的模型选择算法[J].陕西科技大学学报（自然科学版）,2015,(1):179-183,5.基金项目
山西省科技厅科技基础条件平台建设项目 （）
