how can I limit the search for kNN (fitcknn) if I have multiple classes in training but just want to compare the distance to few classes

Hello to all,

how can I limit the classes to few selected.

suppose we have: Bending; Sitting; Standing; FAlling; Walking

in the training

but I just want to compare is it Bending or Walking

while i is the sample and m is the class (five in total) and x is the image

mdl_MD = fitcknn(P_train,movs(trainIdx),'NumNeighbors',3,'Distance','euclidean');

predict(P_train,P_test)

It computes the distance to all classes but I want the distance just to two as I said without computing a new .

I assume this must be possible with the functions fitcknn or predict, but how.

Thank you in advanced for your help.

Answer 1

[Image Analyst]

Only pass in the columns you want, like

pTrain = movs(:, [1,4]); % Extract only bending and walking columns.
testData = movs(:, [1,4]); % Extract only bending and walking columns.
mdl_MD = fitcknn(pTrain,testData,'NumNeighbors',3,'Distance','euclidean');