ExNRule {ExNRuleEnsemble}

The Extended Neighbourhood Rule for k Nearest Neibour Ensemble

Description

The function ExNRule() gives predictions and class probabilities for binary classification data using the extended neighbourhood rule (ExNRule) for k nearest neibour (kNN) ensemble.

Usage

ExNRule(xtrain, xtest, ytrain, k=3, r=500, p=round(sqrt(ncol(xtrain))))

Arguments

xtrain

The features space of the training dataset.

xtest

The feature space of the testing data.

ytrain

The response variable of the training dataset.

k

The nearest observation in the extended pattern.

r

The total number of base kNN models using ExNRule.

p

The number of features used for each base kNN model using ExNRule.

Value

class

Predicted classes for the test data.

class.prob

Predicted class probabilities for the test data.

Author(s)

Amjad Ali, Muhammad Hamraz, Saeed Aldahmani, Zardad Khan.

Maintainer: Amjad Ali <[e-mail]>

References

Ali, Amjad, et al. "An Optimal k Nearest Neighbours Ensemble for Classification Based on Extended Neighbourhood Rule with Features subspace." arXiv preprint arXiv:2211.11278 (2022).

Examples

data(ILPD)

X <- ILPD[,-11]
Y <- as.factor(ILPD[,11])

n <- nrow(ILPD)
train <- sample(1:n, 0.7*n, replace = FALSE)

xtrain <- X[train,]
xtest <- X[-train,]

ytrain <- Y[train]
ytest <- Y[-train]

mymod <- ExNRule(xtrain, xtest, ytrain, k=3, r=500, p=round(sqrt(ncol(xtrain))))
pred.class <- mymod$class
class.prob <- mymod$class.prob

Accuracy <- mean(pred.class==ytest)
ytestnum <- as.numeric(as.factor(ytest))-1
BrierScore <- mean((ytestnum-class.prob)^2)

cbind(Accuracy, BrierScore)