机器学习与数据挖掘 之 模型评估model evalution的方法(Holdout & K-fold Cross-validation & Bootstrap)

我们选择了分类器，训练了模型，那么评价模型/分类器好坏的方法有哪些呢

 

Holdout

Reserve some data for testing

就是划分训练集和测试集

 

K-fold Cross-validation

Divide samples to 푘 partitions, run classifier using 푘 − 1 partitions and test with the remaining
one. Repeat the process for all combinations of 푘 − 1 partitions

 

Bootstrap

Sample n instances with replacement as the training set, use those that is not sampled as the
testing set

This could be done using the Bagging classifier (will not be covered in tutorial)