To use the "fitcknn" in place of "knnclassify" you would have to follow the same procudure as using "knnclassify()", except, you are now training a "model" instead of directly predicting the results. The added advantage to this is that, your model can now be used to predict on unknown data sets too. As a final step after training the model, you have to use the "predict()" function to predict the values on your test dataset. The following code has been modified to accomodate "fitcknn" in place of "knnclassify".
data{1,3} = num2str(RangeR); data{2,3} = num2str(RangeG); data{3,3} = num2str(RangeB); data{4,3} = num2str(RangeH); data{5,3} = num2str(RangeS); data{6,3} = num2str(RangeI);
set(handles.uitable2, 'Data', data, 'ForegroundColor', [0 0 0]);
training1 = xlsread('Data Training');
group = training1(:, 25);
training = training1(:, 1:9);
Z = [MeanR MeanG MeanB MeanH MeanS MeanI VarRed VarGreen VarBlue VarH VarS VarI RangeR RangeG RangeB RangeH RangeS RangeI];
Mdl = fitcknn(training, group, 'NumNeighbors', 3); % example using 3 neighbors, adjust as needed
hasil1 = predict(Mdl, Z);
if hasil1 == 1
x = 'MATURE';
elseif hasil1 == 2
x = 'HALF-MATURE';
elseif hasil1 == 3
x = 'IMMATURE';
end
set(handles.edit2, 'string', x);
You may refer to the documentation below for more information on the mentioned functions:
Hope this helps!
