Holdout and cross-validation strategies for evaluating classifiers.
Interface Summary Interface Description HoldoutA holdout procedure is to provide a range of partitions of a database to pairs of training and test data sets.
Class Summary Class Description AbstractHoldoutSplit a data set for holdout evaluation. DisjointCrossValidationDisjointCrossValidation provides a set of partitions of a database to perform cross-validation. DisjointCrossValidation.ParParameterization class LeaveOneOutA leave-one-out-holdout is to provide a set of partitions of a database where each instances once hold out as a test instance while the respectively remaining instances are training instances. RandomizedCrossValidationRandomizedCrossValidation provides a set of partitions of a database to perform cross-validation. RandomizedCrossValidation.ParParameterization class RandomizedHoldoutA holdout providing a seed for randomized operations. RandomizedHoldout.ParParameterization class StratifiedCrossValidationA stratified n-fold crossvalidation to distribute the data to n buckets where each bucket exhibits approximately the same distribution of classes as does the complete data set. StratifiedCrossValidation.ParParameterization class TrainingAndTestSetWrapper to hold a pair of training and test data sets.