how to make a knn classifer using minkowski distance function

1 次查看（过去 30 天）

Apurva Jariwala 2019-3-17

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/450598-how-to-make-a-knn-classifer-using-minkowski-distance-function

评论： Apurva Jariwala 2019-3-19

Need to make a knn classifer without using fitcknn for K = 3, 5, 7, that uses minkowski distance for the order of 1, 2 and 5

2 个评论
显示无隐藏无

Walter Roberson 2019-3-17

Do you mean that you have been given an assignment to write knn classification code yourself?

If so then it would defeat the purpose if we were to give you knn classification code.

Apurva Jariwala 2019-3-19

在 MATLAB Online 中打开

I am trying to make a knn classifier and train and test it using the Iris dataset. The objective is to find accuracy and the confusion matrix. Please read the code below and let me know what changes can I make

IrisD = readtable('irisdata.csv');
classes = categorical(IrisD{:,5});
Icats = categories(classes);
setosa = IrisD(strcmp(IrisD{:,5},Icats(1)),:);
Ttest1 = setosa(1:40,:);
Ttrain1 = setosa(41:50,:);
versicolor = IrisD(strcmp(IrisD{:,5},Icats(2)),:);
Ttest2 = versicolor(1:40,:);
Ttrain2 = versicolor(41:50,:);
virginica = IrisD(strcmp(IrisD{:,5},Icats(3)),:);
Ttest3 = virginica(1:40,:);
Ttrain3 = virginica(41:50,:);
Ttest = [Ttest1; Ttest2; Ttest3];
Ttrain = [Ttrain1; Ttrain2; Ttrain3];
testlabel = Ttest(:,5);
trainlabel = Ttrain(:,5);
C = unique(trainlabel);
testf = Ttest(:,1:4);
trainf = Ttrain(:,1:4);
K = 3;
r = 2;
Lpred = [];
for i = 1:size(testf,1)
    Ftest = testf(i,:);
    Ns = size(trainf, 1);
    dmat = abs(trainf-repmat(Ftest, Ns, 1));
    dlist = nthroot(sum(dmat.^r, 2), r);
    [dsort, isort] = sort(dlist, 'ascend');
    Lknn = trainlabel(isort(1:K));
    Ncl = []; 
    for iC = 1:length(C) 
        cl = C(iC);
        ncl = length(find(Lknn==cl)); 
        Ncl = [Ncl; ncl, cl];
    end
    [vmax, imax] = max(Ncl(:,1));
    Cpred = Ncl(imax, 2);
    Lpred = [Lpred; Cpred];
end