Therefore, a subset of the available test cases is used for training and a disjoint
subset is used for validation. In cross validation, the available test cases are
divided into k parts numbered 1 to k, fromwhich k different test sets are created
as follows: test set i uses the ith part for validation, after training the classifier
using the other k−1parts.The results (t pos, f pos, etc.) fromall k test sets are added
up before computing the quality measures. Cross validation provides much more
accurate measures than merely partitioning the data into a single training and a
single test set