splitEqual: splitEqual

In Mthrun/Classifiers: Common Supervised Machine Learning Algorithms

splitEqual

Description

Equal split of Data randomly into disjunct sets TrainCls and TestCls, with |TrainCls| = |TestCls|*Percentage/100. Splitting is done such that given a percentage the cases over all classes are the same. Data can be also divided into disjunct set in this function otherwise use indices to split Data in a later step.

Usage

splitEqual(Cls, Data, Percentage, ForceEqual = FALSE, Key)

Arguments

Cls	[1:n] Numeric vector of classifications, with k unique classes
Data	Optional, [1:n,1:d] Matrix of n cases and d features, or n d-dimensional datapoints
Percentage	Scalar, gives the percentage for the split as value between 1 and 99. If missing default is 50
ForceEqual	TRUE: sample with replacement if percentage of full cases is above the number of cases in a class, default: FALSE: less cases are sampled resulting in an still unequal split If a class has a low number of cases
Key	Optional, [1:n] vector, containing the keys. Default is either 1:length(Cls) or rownames(Data) if existing

Details

Sample without replacement is performend in the default case.

TrainInd and TestInd are permutated.

Value

List with,

TrainInd	Numeric vector, containing the indices of the the Cls/Data entries in the Train split
TestInd	Numeric vector, containing the indices of the the Cls entries in the Test split
TrainCls	Numeric vector, containing the cls entries in the Train split
TestCls	Numeric vector, containing the cls entries in the Test split
TrainData	Numeric vector, containing the Data entries in the Train split. If Data is missing, then TrainData is NULL
TestData	Numeric vector, containing the Data entries in the Test split. If Data is missing, then TestData is NULL
TrainKey	Numeric vector, containing the key entries in the Train split
TestKey	Numeric vector, containing the key entries in the Test split

Author(s)

Michael Thrun

Examples

data("USelectionPoll")
V=splitEqual(USelectionPoll$Cls,Percentage=50)

Mthrun/Classifiers documentation built on June 28, 2023, 9:28 a.m.