pca_fuzzy_c-means_method.txt

首先对原始数据 归一化 然后进行PCA分析 采用PCs作为特征

load ushape.dat
%perfomring PCA
[pcs,newdata,variances,t2] = princomp(ushape);
percent_explained = 100*variances/sum(variances)
%plot Pareto
subplot(3,3,1)
pareto(percent_explained)
title('Pareto plot')
xlabel('Principal Component')
ylabel('Variance Explained (%)')

%plot PC1-PC2 
subplot(3,3,2)
line(newdata(:, 1), newdata(:, 2), 'linestyle',...
'none','marker', 'd','color','b');
title('PC1-PC2 two dimension plot')
xlabel('PC1')
ylabel('PC2')

%fuzzy c-means clustering
n=input('Enter the number of clusters wanted:')
m=input('Enter the number of PCs are wanted:')
[center,U,objFcn] =fcm(newdata(:,1:m),n);

maxU = max(U);
for k=1:1:n
index(k) = {find(U(k, :) == maxU)};
end

subplot(3,3,3)

linecolor='rgbcmyk';
pointmark='dox+*sv^<>ph';

for k=1:1:n
line(newdata(index{k}, 1), newdata(index{k}, 2), 'linestyle','none','marker', pointmark(k),'color',linecolor(k));
end

hold on
for k=1:1:n
plot(center(k,1),center(k,2),strcat('k',pointmark(k)),'markersize',15,'LineWidth',2)
end

title('Fuzzy c-means clustering')
xlabel('PC1')
ylabel('PC2')

[row1,col1]=size(ushape)
for k=1:1:n
    [row2,col2]=size(index{k})
    cluster(k)={[]}
    for p=1:1:col2
        cluster(k)={[cluster{k};ushape(index{k}(p),1:col1)]}
    end
end
%
%for k=1:1:n
%    subplot(3,3,3+k)
%    hist (cluster{k},3)
%    xlabel('Springback Angle')
%    %ylabel('normal probability density')
%    title(strcat('cluster',int2str(k)))
%end