knncTrain

Training of KNNC (K-nearest neighbor classifier)

Contents

Syntax

Description

knncPrm = knncTrain(DS, knncTrainPrm, plotOpt) returns the parameters of KNNC after training, where

• DS: the training set
• knncTrainPrm: parameters for training, including
  • knncTrainPrm.method: 'none', 'kMeans', or 'kMeans&lvq'
  • knncTrainPrm.centerNum4eachClass: no. of prototypes for each class
• knncPrm: parameters for KNNC, including the following necessary fields:
  • knncPrm.method: method for training
    • 'none': none
    • 'kMeans': k-means clustering for training
  • knncPrm.k: the value of k in KNNR
  • knncPrm.class: class parameters
    • knncPrm.class(i).data: prototypes (centers) for class i

Example

knncTrainOpt=knncTrain('defaultOpt');
knncTrainOpt.method='kMeans';
knncTrainOpt.centerNum4eachClass=4;
[trainSet, testSet]=prData('3classes');
[knncPrm, logLike, recogRate]=knncTrain(trainSet, knncTrainOpt, 1);
fprintf('Inside recog. rate=%.2f%%\n', recogRate*100);
cClass=knncEval(testSet, knncPrm);
hitIndex=find(cClass==testSet.output);
recogRate=length(hitIndex)/length(cClass);
fprintf('Outside recog. rate=%.2f%%\n', recogRate*100);
testSet.hitIndex=hitIndex;
figure; knncPlot(testSet, knncPrm, 'decBoundary');

Inside recog. rate=96.00%
Outside recog. rate=92.00%

Published with MATLAB® R2018a