Question

我实现了K-Nearest Neighbors算法，但我使用matlab的经验很少。我需要你检查一小部分代码并告诉我可以改进或修改什么？并希望它是算法的正确实现？

function test_data = knn(test_data, tr_data,k)

numoftestdata = size(test_data,1);
numoftrainingdata = size(tr_data,1);


for sample=1:numoftestdata

   %Step 1: Computing euclidean distance for each testdata
   R = repmat(test_data(sample,:),numoftrainingdata,1) ;
   euclideandistance  = (R(:,1) - tr_data(:,1)).^2;

   %Step 2: compute k nearest neighbors and store them in an array
    [dist position] = sort(euclideandistance,'ascend');
    knearestneighbors=position(1:k);
    knearestdistances=dist(1:k);


    % Step 3 : Voting 
    for i=1:k
        A(i) = tr_data(knearestneighbors(i),2);  
    end

    M = mode(A);

    if (M~=1)
        test_data(sample,2) = mode(A);
    else 
        test_data(sample,2) = tr_data(knearestneighbors(1),2);
    end
end

要测试它，您可以使用：

test_data = [6,0; 2,0; 5,0]
tr_data = [1,1; 0,2; 3,2; 4,4; 5,3]

Answer 1

我会做以下更正：

第3步：循环效率不高，请尝试A =tr_data(knearestneighbors,2);
if语句可以简化

-

test_data(sample,2) = mode(A);

if (M==1)
    test_data(sample,2) = tr_data(knearestneighbors(1),2);
end

这个Matlab代码可以改进什么？这是KNN的正确实施吗？

1 个答案: