knncEval

K-nearest neighbor classifier (KNNC)

Contents

Syntax

Description

computedClass = knncEval(testSet, knncPrm) returns the results of KNNC, where

• testSet: the test set
• knncPrm: parameters for KNNC
  • knncPrm.k: the value of k for k-nearest neighbor classification
• computedClass: output vector by KNNC

[computedClass, logLike, recogRate hitIndex, nnIndex] = knncEval(testSet, knncPrm) returns extra info:

• testSet: the test set
• logLike: log likelihood for each of the test set
• recogRate: recognition rate
• hitIndex: index of correctly classified test pairs
• nnIndex: index of knncPrm.input that are closest to testSet.input

Example

[trainSet, testSet]=prData('iris');
trainNum=size(trainSet.input, 2);
testNum =size(testSet.input, 2);
fprintf('Use of KNNC for Iris data:\n');
fprintf('\tSize of train set (odd-indexed data) = %d\n', trainNum);
fprintf('\tSize of test set (even-indexed data) = %d\n', testNum);
knncTrainPrm.method='none';
knncPrm=knncTrain(trainSet, knncTrainPrm);
fprintf('\tRecognition rates as K varies:\n');
kMax=15;
for k=1:kMax
	knncPrm.k=k;
	computed=knncEval(testSet, knncPrm);
	correctCount=sum(testSet.output==computed);
	recogRate(k)=correctCount/testNum;
	fprintf('\t%d-NNC ===> RR = 1-%d/%d = %.2f%%.\n', k, testNum-correctCount, testNum, recogRate(k)*100);
end
plot(1:kMax, recogRate*100, 'b-o'); grid on;
title('Recognition rates of Iris data using KNN classifier');
xlabel('K'); ylabel('Recognition rates (%)');

Use of KNNC for Iris data:
	Size of train set (odd-indexed data) = 75
	Size of test set (even-indexed data) = 75
	Recognition rates as K varies:
	1-NNC ===> RR = 1-3/75 = 96.00%.
	2-NNC ===> RR = 1-4/75 = 94.67%.
	3-NNC ===> RR = 1-3/75 = 96.00%.
	4-NNC ===> RR = 1-3/75 = 96.00%.
	5-NNC ===> RR = 1-2/75 = 97.33%.
	6-NNC ===> RR = 1-2/75 = 97.33%.
	7-NNC ===> RR = 1-2/75 = 97.33%.
	8-NNC ===> RR = 1-1/75 = 98.67%.
	9-NNC ===> RR = 1-4/75 = 94.67%.
	10-NNC ===> RR = 1-2/75 = 97.33%.
	11-NNC ===> RR = 1-2/75 = 97.33%.
	12-NNC ===> RR = 1-3/75 = 96.00%.
	13-NNC ===> RR = 1-3/75 = 96.00%.
	14-NNC ===> RR = 1-2/75 = 97.33%.
	15-NNC ===> RR = 1-2/75 = 97.33%.

Published with MATLAB® R2018a