I want to use fitcknn but to iterate for multiple k variables ( k = 1:15) and then store the accuracy values in a matrix. What is an easy way to do this? Thanks!

clear all
data = readtable('some_data.txt');

%%%% Normalize Data %%%%
x = (data.x - mean(data.x))/std(data.x);
data.x = x;

y = (data.y - mean(data.y))/std(data.y);
data.y = y;

k = 15

%%%% Build Classifier %%%%
    class_model = fitcknn(data,'c~x+y','NumNeighbors',k,'Distance','euclidean');

%%%% Partition Training/Testing sets %%%%
    cv = cvpartition(class_model.NumObservations,'HoldOut',0.2);
    cross_val_model = crossval(class_model,'cvpartition',cv);

%%%% Create Predictions %%%%
    Predict = predict(cross_val_model.Trained{1},data(test(cv),1:end-1));

%%%% Draw Confusion Matrix %%%%
    Result = confusionmat(cross_val_model.Y(test(cv)),Predict);

%%% Accuracy %%%%
    accuracy = ((Result(2,2)+Result(1,1))/((Result(2,2)+Result(1,1)+Result(1,2)+Result(2,1))))
    

[Nicholas Boselli]

I think I just did it but as a beginner my code is still sloppy so if you have a "clean" way of doing this please share! Thanks all!

Answer 1

[Walter Roberson]

clear all
data = readtable('some_data.txt');

%%%% Normalize Data %%%%
x = (data.x - mean(data.x))/std(data.x);
data.x = x;

y = (data.y - mean(data.y))/std(data.y);
data.y = y;

Kvals = 1:15;

%%%% Build Classifier %%%%
    class_models = arrayfun(@(k) fitcknn(data,'c~x+y','NumNeighbors',k,'Distance','euclidean'), Kvals, 'uniform', 0);

%%%% Partition Training/Testing sets %%%%
    cvs = cellfun(@(CM) cvpartition(CM.NumObservations,'HoldOut',0.2), class_models, 'uniform', 0);
    cross_val_models = cellfun(@(CM,CV) crossval(CM,'cvpartition',CV), class_models, cvs, 'uniform', 0);

%%%% Create Predictions %%%%
    Predicts = cellfun(@(CVM, CV) predict(CVM.Trained{1},data(test(CV),1:end-1)), cross_val_models, cvs, 'uniform', 0);

%%%% Draw Confusion Matrix %%%%
    Results = cellfun(@(CVM, CV, P) confusionmat(CVM.Y(test(CV)),P), cross_val_models, cvs, Predicts, 'uniform', 0);

%%% Accuracy %%%%
    accuracys = cellfun(@(Result) ((Result(2,2)+Result(1,1))/((Result(2,2)+Result(1,1)+Result(1,2)+Result(2,1)))), Results);
    
    plot(Kvals, accuracys)