My K-means program has malfunctionality; please help me to run it properly

Question

Federico Frics 2015-4-12

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/197841-my-k-means-program-has-malfunctionality-please-help-me-to-run-it-properly

编辑： Federico Frics 2015-4-13

采纳的回答： Federico Frics

fisheriris.m

在 MATLAB Online 中打开

Hello everyone,

I wrote some scripts for K-means clustering, but sometimes it is not able to assign some points to its correct cluster. As it seems, there are some problems that I'm not to find them and revise them. I put my scripts here and need your idea to rearrange the program for running appropriately. In addition, the data of these codes were attached, and you can find it at the bottom of this post.

clc;
clear;
close all;
load fisheriris
ClustersNumber=2;
X=meas(:,3);
Y=meas(:,4);
Random=randi([1 numel(X)],ClustersNumber,1);

for i=1:ClustersNumber

     Mean.X(i)=X(randi(Random(i)));
     Mean.Y(i)=Y(randi(Random(i)));

end

iteration=0;

while 1

% The distance of each point from the mean of each cluster

    if iteration>100
        for j=1:numel(X)
            for i=1:ClustersNumber
            Cluster.Distance(i,j)=sqrt((X(j)-Mean.X(i))^2+(Y(j)-Mean.Y(i))^2);
              Cluster.Distance(Cluster.Distance==0)=NaN;
            end
             Min(j)=min(Cluster.Distance(:,j));
         end
    %%Assigning the points to each nearest cluster respecting the means
    for i=1:numel(X)
            if Cluster.Distance(1,i)==Min(i)
                Class(1).data(i)=Min(i);
                Class(1).X(i)=X(i);
                Class(1).Y(i)=Y(i);
            else
               Class(2).data(i)=Min(i);
                Class(2).X(i)=X(i);
                Class(2).Y(i)=Y(i);
            end  
    end
    % Updating the means
    for i=1:ClustersNumber
         Mean.X(i)=mean(Class(i).X(Class(i).X~=0));
         Mean.Y(i)=mean(Class(i).Y(Class(i).Y~=0));
    end
    break;
    end
 iteration=iteration+1;

end

scatter(Class(1).X,Class(1).Y,'+'); hold on; scatter(Class(2).X,Class(2).Y,'s');

Thanks

2 个评论
显示无隐藏无

Image Analyst 2015-4-12

在 MATLAB Online 中打开

All I get is this:

>> load fisheriris
Error using load
Unable to read file 'fisheriris'. No such file or directory.

How do you think we can solve that so we can help you? Please read this.

Federico Frics 2015-4-13

编辑：Federico Frics 2015-4-13

Thanks for your response. The fisheriris is a data incorporated into matlab directory by default. However, if you did'nt manage to find the data, I attached the data to the first post for you.

Thanks

请先登录，再进行评论。

请先登录，再回答此问题。

Answer 1

Federico Frics 2015-4-13

0
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/197841-my-k-means-program-has-malfunctionality-please-help-me-to-run-it-properly#answer_175034

编辑：Federico Frics 2015-4-13

在 MATLAB Online 中打开

Hi there,

As no one was not able to help me, I found where the problem is, and revised it.Therefore, I place my final codes here to help other people.

clc; clear; close all; load fisheriris ClustersNumber=2; X=meas(:,3); Y=meas(:,4); Random=randi([1 numel(X)],ClustersNumber,1);

 %%Generate the initial clusters randomly
 Class(1).X=X(1:(randi([1 80],1)));
 Class(1).Y=Y(1:numel(Class(1).X));
 Class(2).X=X(numel(Class(1).X)+1:end);
 Class(2).Y=Y(numel(Class(1).X)+1:end);

% Determine the mean of each initial cluster for k=1:ClustersNumber

     Mean.X(k)=mean(Class(k).X);
     Mean.Y(k)=mean(Class(k).Y);
     end

%% Main loop iteration=0;

while 1

    if iteration>100
        % The distance of each point from the mean of the clusters 
        for j=1:numel(X)
            for i=1:ClustersNumber
            Cluster.Distance(i,j)=sqrt((X(j)-Mean.X(i))^2+(Y(j)-Mean.Y(i))^2);
              Cluster.Distance(Cluster.Distance==0)=NaN;
            end
            % Compute the minimum distance of each point from the classes
             Min(j)=min(Cluster.Distance(:,j));
         end
    %%Assign the points to each nearest cluster respecting the means
    for i=1:numel(X)
            if Cluster.Distance(1,i)==Min(i)
                NewClass(1).data(i)=Min(i);
                NewClass(1).X(i)=X(i);
                NewClass(1).Y(i)=Y(i);
            elseif Cluster.Distance(2,i)==Min(i)
               NewClass(2).data(i)=Min(i);
                NewClass(2).X(i)=X(i);
                NewClass(2).Y(i)=Y(i);
            end  
    end
    % Update the means
        for k=1:ClustersNumber
            NewClass(k).X( NewClass(k).X==0)=[];
            NewClass(k).Y( NewClass(k).Y==0)=[];
            Mean.X(k)=mean(NewClass(k).X);
            Mean.Y(k)=mean(NewClass(k).Y);
        end
    break;
    end
 iteration=iteration+1;

end

scatter(NewClass(1).X,NewClass(1).Y,'+'); hold on; scatter(NewClass(2).X,NewClass(2).Y,'s'); scatter(Mean.X(1),Mean.Y(1),'*'); scatter(Mean.X(2),Mean.Y(2));