cluster.m

进行基于最大最小方法和K均值结合的分类方法 并画出具体的分类情况

function [ p_cluster] = cluster(p_sample, k, feature_num, sample_num)
% Detailed explanation goes here, feature_num and sample_num are
% dimensions of the sample，k represents the ideal classified numbers 
load('test.mat');
%initialize the k centres;
p_cluster = max_min(p_sample);
p_cluster = p_cluster';%(p_cluster-500*51)

%iterative process
for i=1:500
    p_dist = dist(p_cluster, p_sample);
    [Y, I] = min(p_dist);
    clear Y;
    %recomputering the centers
    for j=1:k
        sum = zeros(feature_num,1);
        count = 0;
        for m=1:sample_num
            if(I(m) == j)
                sum = sum + p_sample(:,m);
                count=count+1;
            end
        end
        if(count > 0)
            sum = sum/count;
            p_cluster(j,:) = sum';
        end
    end
end
for m=1:sample_num
    if(I(m)==1)
    plot(p_sample(1,m),p_sample(2,m),'ro');
    end
    hold on;
    if(I(m)==2)
    plot(p_sample(1,m),p_sample(2,m),'yo');
    end
    hold on 
    if(I(m)==3)
    plot(p_sample(1,m),p_sample(2,m),'go');
    hold on 
    end
end
radius= zeros(1,k);
for j=1:k
      sub = find(I == j);
      p_max= dist(p_cluster(j,:),p_sample(:,sub));
      [Y,M]= max(p_max);
      clear M;
      radius(j)=Y;
end
p_cluster = p_cluster';
plot(p_cluster(1,:),p_cluster(2,:),'b*');
hold on;
y=p_cluster(2,:);
x=p_cluster(1,:);
for i=1:k    
plotcircle(x(i),y(i),radius(i)+0.5);
end
hold on;
bisector(p_cluster(:,1),p_cluster(:,3));
hold on;
bisector(p_cluster(:,1),p_cluster(:,2));
hold on;
bisector(p_cluster(:,2),p_cluster(:,3));
hold on;
axis([0 35 0 25]);
end