Kmeans算法matlab的实现Word下载.docx
《Kmeans算法matlab的实现Word下载.docx》由会员分享,可在线阅读,更多相关《Kmeans算法matlab的实现Word下载.docx(16页珍藏版)》请在冰点文库上搜索。
Data(51:
100,1)=MH;
50,2)=FW;
100,2)=MW;
C=input('
Ê
ä
È
ë
C£
º
'
)
[U,P,Dist,Cluster_Res,Obj_Fcn,iter]=fuzzycm(Data,C)
plot(Data(:
1),Data(:
2),'
o'
holdon;
maxU=max(U);
index1=find(U(1,:
)==maxU);
index2=find(U(2,:
line(Data(index1,1),Data(index1,2),'
marker'
*'
color'
g'
line(Data(index2,1),Data(index2,2),'
r'
plot([P([12],1)],[P([12],2)],'
k'
)
holdoff;
2、子程序
function[U,P,Dist,Cluster_Res,Obj_Fcn,iter]=fuzzycm(Data,C,plotflag,M,epsm)
ifnargin<
5
epsm=1.0e-6;
end
4
M=2;
3
plotflag=0;
[N,S]=size(Data);
m=2/(M-1);
iter=0;
Dist(C,N)=0;
U(C,N)=0;
P(C,S)=0;
U0=rand(C,N);
U0=U0./(ones(C,1)*sum(U0));
whiletrue
iter=iter+1;
Um=U0.^M;
P=Um*Data./(ones(S,1)*sum(Um'
))'
;
fori=1:
C
forj=1:
N
Dist(i,j)=fuzzydist(P(i,:
),Data(j,:
));
end
end
U=1./(Dist.^m.*(ones(C,1)*sum(Dist.^(-m))));
ifnargout>
4|plotflag
Obj_Fcn(iter)=sum(sum(Um.*Dist.^2));
ifnorm(U-U0,Inf)<
epsm
break
U0=U;
ifnargout>
3
res=maxrowf(U);
forc=1:
v=find(res==c);
Cluster_Res(c,1:
length(v))=v;
ifplotflag
fcmplot(Data,U,P,Obj_Fcn);
function[U,P,Dist,Cluster_Res,Obj_Fcn,iter]=fuzzycm2(Data,P0,plotflag,M,epsm)
[N,S]=size(Data);
m=2/(M-1);
iter=0;
C=size(P0,1);
U(C,N)=0;
P(C,S)=0;
Dist(i,j)=fuzzydist(P0(i,:
Um=U.^M;
ifnorm(P-P0,Inf)<
P0=P;
functionf=addr(a,strsort)
ifnargin==1
strsort='
ascend'
sa=sort(a);
ca=a;
la=length(a);
f(la)=0;
fori=1:
la
f(i)=find(ca==sa(i),1);
ca(f(i))=NaN;
ifstrcmp(strsort,'
descend'
f=fliplr(f);
functionellipse(a,b,center,style,c_3d)
style='
b'
3|isempty(center)
center=[0,0];
t=1:
360;
x=a/2*cosd(t)+center
(1);
y=b/2*sind(t)+center
(2);
ifnargin>
plot3(x,y,ones(1,360)*c_3d,style)
else
plot(x,y,style)
functionfcmplot(Data,U,P,Obj_Fcn)
[C,S]=size(P);
str='
po*x+d^v>
<
.h'
figure
(1),plot(Obj_Fcn)
title('
¿
±
ê
¯
ý
Ö
µ
»
Ç
ú
Ï
ß
fontsize'
8)
ifS==2
figure
(2),plot(P(:
1),P(:
rs'
),holdon
v=Data(find(res==i),:
plot(v(:
1),v(:
2),str(rem(i,12)+1))
ellipse(max(v(:
1))-min(v(:
1)),...
max(v(:
2))-min(v(:
2)),...
[max(v(:
1))+min(v(:
2))+min(v(:
2))]/2,'
r:
gridon,title('
2D¾
Û
À
à
á
¹
û
Í
8),holdoff
ifS>
2
figure
(2),plot3(P(:
2),P(:
3),'
plot3(v(:
2),v(:
3),str(rem(i,12)+1))
2))]/2,...
'
(max(v(:
3))+min(v(:
3)))/2)
3D¾
functionD=fuzzydist(A,B)
D=norm(A-B);
functionmr=maxrowf(U,c)
2
c=1;
N=size(U,2);
mr(1,N)=0;
forj=1:
aj=addr(U(:
j),'
mr(j)=aj(c);
四、实验结果
1、FEMALE和MALE
(1)C=2,Z1
(1)=(173,53)T,Z2
(1)=(168,57)T。
聚类中心为(163.322052.5232),(175.256567.6907)
迭代次数为28
(2)C=2,Z1
(1)=(173,53)T,Z2
(1)=(160,58)T。
(3)C=3,Z1
(1)=(173,53)T,Z2
(1)=(168,57)TZ3
(1)=(160,58)T
聚类中心为(168.413957.3300)(176.335169.4859)
(160.176749.1940)
迭代次数为41
(4)C=3,Z1
(1)=(173,53)T,Z2
(1)=(168,57)TZ3
(1)=(161,45)T
迭代次数为41
2、test2
聚类中心为(166.844354.9270),(176.677270.7966)
迭代次数为53
聚类中心为(178.869980.0977)(163.946951.6216)
(174.814464.8963)
迭代次数为56
(174.814464.8963)
迭代次数为56
结论
(一):
初始均值设置的不同会影响迭代的次数以及各次迭代所产生的聚类中心,但它不会影响最后的分类结果。
结论
(二):
数据的输入顺序不同,同样影响迭代次数,而对聚类结果没有太大的影响。