leave-one-out using fitcdiscr

As classify will soon disappear, I am motivated to learn fitcdiscr. I prefer leave-one-out crossval and typically use classify in the following manner to achieve leave-one-out:

load fisheriris %gives meas for sample, and species

group = NaN(150,1); %Create group to use instead of species

group(1:50)=1;

group(51:100)=2;

group(101:150)=3;

totalMeasurements = length(meas);

predClass = NaN(length(meas),1);

for i = 1:totalMeasurements;

%assemble training and test sets for loop

testingData = meas(i,:);

testingLabel = group(i);

%take whle data as training set

trainingData = meas;

trainingLabel = group;

%NaN out test sample

trainingData(i,:) = NaN;

trainingLabel(i) = NaN;

%remove NaNs from training set

bad = isnan(trainingLabel);

trainingLabel = trainingLabel(~bad);

trainingData = trainingData(~bad,:);

%use matlab classify function

predClass(i)= classify(testingData, trainingData, trainingLabel);

end

correct = predClass ==group;

percent_correct = sum(correct)/length(correct);

How do I use fitcdiscr to replace this? I've fooled around a bit with 'Leaveout' but I can't find good examples to follow. Thanks for your time!

Answer 1

[Pratik]

Hi Lauren,

The example given in 'fitcdiscr' shows the usage of this function over the 'fisheriris' dataset.

Please refer to the following documentation of same:

www.mathworks.com/help/stats/fitcdiscr.html#bt6d86x-5