how to solve the array index problem in kmeans clustering for three dimensional data?

2 次查看（过去 30 天）

AS 2020-9-17

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/595366-how-to-solve-the-array-index-problem-in-kmeans-clustering-for-three-dimensional-data

编辑： AS 2020-9-17

I have used a with dimension 100 by 3 of 10 signals for kmeans clustering. So, I have used it as a three dimensional dataset of induvidual signal. I have used kmeans clustering but code is not working, showing the error index exceeds the number of array elements (1). Please provide me suggestion to solve out the problem.

niters=200;
Nouter=5;
nc=3; % clunster number
for kk=1:niters
    for ii=1:100
        for jj=1:10
            old_c(:,:,jj)=c(:,:,jj);
            d2(:,:,jj)=dist2(data(:,:,jj),c(:,:,jj));
            [minval,indx]=min(d2(:,:,jj),[],2);
            post1(:,:,jj)=id1(indx,jj);
            num_points1(:,:,jj)= sum(post1(:,:,jj),1);
            for k=1:Nouter
                for n=1:nc
                    if num_points1(n) > 0 
                        c(n,:,jj)=sum(data(find(post1(:,n,jj)),:,jj),1)./num_points1(:,jj);
                    end
                end
                e1(:,:,jj)=sum(minval);
                errlog(kk)=e1(:,:,jj);
                if kk>1
                    if max(max(abs(c(:,:,jj)-old_c(:,:,jj)))) < 0.0001 &&  abs(old_e(:,:,jj)-e1(:,:,jj)) < 0.0001
                        new_e(:,:,jj)=e1(:,:,jj);
                        return;
                    end
                end
                old_e(:,:,jj)=e1(:,:,jj);
            end
            new_e(:,:,jj)=e1(:,:,jj);
        end
    end
end