KODAMA (version 0.0.1)

knn.dist: Calculates the Distances for KNN Predictions

Description

The distances to be used for K-Nearest Neighbor (KNN) predictions are calculated and returned as a symmetric matrix. Distances are calculated by dist.

Usage

knn.dist(x, dist.meth = "euclidean", p = 2)

Arguments

x

a matrix of data.

dist.meth

the distance to be used in calculating the neighbors. Any method valid in function dist is valid.

p

the power of the Minkowski distance.

Value

a square symmetric matrix whose dimensions are the number of rows in the original data. The diagonal contains zeros, the off diagonal entries will be >=0..

Details

This function calculates the distances to be used by knn.predict. Distances are calculated between all cases. In the traditional scenario. The advantage to calculating distances in a separate step prior to prediction, is that these claculations only need to be performed once. So, for example, cross-validation to select k can be performed on many values of k, with different cross-validation splits, all using a single run of knn.dist. The default method for calculating distances is the "euclidean" distance, which is the method used by the knn function from the class package. Alternative methods may be used here. Any method valid for the function dist is valid here. The parameter p may be specified with the Minkowski distance to use the p norm as the distance method.

See Also

knn.predict,dist

Examples

	#a quick classification example
	x1 <- c(rnorm(20, mean=1), rnorm(20, mean=5))
	x2 <- c(rnorm(20, mean=5), rnorm(20, mean=1))
	y=rep(1:2,each=20)
	x <- cbind(x1,x2)
	train <- sample(1:40, 30)
	#plot the training cases
	plot(x1[train], x2[train], col=y[train]+1)
	#predict the other cases
	test <- (1:40)[-train]
	kdist <- knn.dist(x)
	preds <- knn.predict(train, test, y ,kdist, k=3, agg.meth="majority")
	#add the predictions to the plot
	points(x1[test], x2[test], col=as.integer(preds)+1, pch="+")
	#display the confusion matrix
	table(y[test], preds)