用MATLAB实现数据挖掘的一种算法Word文档下载推荐.docx
《用MATLAB实现数据挖掘的一种算法Word文档下载推荐.docx》由会员分享,可在线阅读,更多相关《用MATLAB实现数据挖掘的一种算法Word文档下载推荐.docx(16页珍藏版)》请在冰豆网上搜索。
![用MATLAB实现数据挖掘的一种算法Word文档下载推荐.docx](https://file1.bdocx.com/fileroot1/2022-12/12/619a7fe8-16a8-4b33-a92e-f0cdba48dac4/619a7fe8-16a8-4b33-a92e-f0cdba48dac41.gif)
使用
dm=xlsread('
lisanhua'
);
导入’lisanhua’.xls文件,在MATLAB中对一些连续属性值离散化。
如下:
1、[29,48]=1,[48,62]=2,[62,77]=3
4、[94,110]=1,[110,143]=2,[143,200]=3
5、[126,205]=1,[205,293]=2,[293,564]=3
8、[71,120]=1,[120,175]=2,[175,202]=3
10、[0,1.5]=1,[1.5,2.5]=2,[2.5,6.2]=3。
然后对数据进行归一化处理:
由于不同属性之间的属性值相同,所以利用下面语句对一共13个条件属性中的38个属性进行如下赋值,使每条属性唯一确定。
从而得到38个条件属性,只不过38个里面有且只能出现13个。
程序如下:
m1=[0,3,2,4,3,3,2,3,3,2,3,3,4,3];
k=1;
w=m1(k);
dm3=dm2;
fori=1:
3939
dm3(i)=dm2(i)+w;
ifrem(i,303)==0
k=k+1;
w=w+m1(k);
end
end
从而得到dm(3)矩阵。
而且决策属性分为1:
healthy;
2:
sick。
并且在38个条件属性中没有值为0。
四、挖掘算法
1、特殊样本
首先在前200条样本中分别找出三条对应两种决策属性的重要样本,样本必须满足在同类决策属性下其他199条的13条决策属性中和它的13条条件属性数目大于等于10的前3条样本。
2、特殊条件属性值
其次分别对应两条决策属性值的5条重要条件属性值(在38个条件属性里找),特殊属性值必须满足:
(1)在对应相同决策属性下,此决策属性支持率必须在前五,
(2)而且如果不同决策属性出现相同条件属性。
如果相同决策属性同时出现在不同决策属性中,删除这条后找支持率第六的条件属性,以此类推。
挖掘算法在MATLAB里列出并做了标注。
五、验证程序
1、预处理
在验证程序里面分别对测试数据和六条样本做了对比,又对其属性值和特殊属性值做了对比,最后利用加权求和算法判断测试样本的决策属性正确率。
其中:
nc,mc代表测试数据分别和两类样本属性中相同数是否大于等于9时的加权值。
ncc,mcc代表测试数据分别和两类特殊属性满足几条数目的加权值。
2、关联规则
(1)、如果(nc>
=0.9&
ncc>
=8)|(ncc>
=10&
nc>
=0.6)得到决策属性healthy
(2)、不满足以上条件的话如果(mc>
=0.9|mcc>
=8)|(mcc>
=9&
mc>
=0.6)得到决策属性sick。
(3)、不满足以上条件的话如果nc>
=(mc+0.3)|ncc>
=(mcc+2)得到决策属性healthy。
(4)、不满足以上条件的话如果mc>
nc得到决策属性为sick。
(5)、不满足以上条件的话如果ncc>
mcc得到决策属性为healthy。
(6)、不满足以上条件的话如果mcc>
ncc得到决策属性为sick。
(7)、不满足以上条件的话得到决策属性为sick。
六、正确率及结果分析
正确率为82.5%
结果分析:
由于采用了两类约束方法,所以效果还可以。
七、程序如下页所示
clear;
%%%%%数据预处理程序%%%%%%
%载入数据
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
z1=dm(:
1);
%离散化第一列数据
303
ifz1(i)>
=29&
z1(i)<
48
z1(i)=1;
elseifz1(i)>
=48&
62
z1(i)=2;
elsez1(i)=3;
z2=dm(:
2);
ifz2(i)==0
z2(i)=1;
elsez2(i)=2;
z3=dm(:
3);
z3(i)=z3(i)+1;
z4=dm(:
4);
ifz4(i)>
=94&
z4(i)<
110
z4(i)=1;
elseifz4(i)>
=110&
143
z4(i)=2;
elsez4(i)=3;
z5=dm(:
5);
ifz5(i)>
=126&
z5(i)<
205
z5(i)=1;
elseifz5(i)>
=205&
293
z5(i)=2;
elsez5(i)=3;
z6=dm(:
6);
z6(i)=z6(i)+1;
z7=dm(:
7);
z7(i)=z7(i)+1;
z8=dm(:
8);
ifz8(i)>
=71&
z8(i)<
120
z8(i)=1;
elseifz8(i)>
=120&
175
z8(i)=2;
elsez8(i)=3;
z9=dm(:
9);
z9(i)=z9(i)+1;
z10=dm(:
10);
ifz10(i)>
=0&
z10(i)<
1.5
z10(i)=1;
elseifz10(i)>
=1.5&
2.5
z10(i)=2;
elsez10(i)=3;
z11=dm(:
11);
z11(i)=z11(i)+1;
z12=dm(:
12);
z12(i)=z12(i)+1;
z13=dm(:
13);
z13(i)=z13(i)+1;
z14=dm(:
14);
z14(i)=z14(i)+1;
dm2=[z1,z2,z3,z4,z5,z6,z7,z8,z9,z10,z11,z12,z13,z14];
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%%%%%预处理结束%%%%%%
%%%%挖掘算法%%%%%%
%%%%%%%%%%%%%%%%%%%%%%%找6条样本前的预处理
dm4=zeros(200,14);
dm5=zeros(200,14);
200
ifdm3(i,14)==1
dm4(i,1:
13)=dm3(i,1:
13);
elsedm5(i,1:
a1=zeros(38,1);
a2=zeros(38,1);
fork=1:
38
13
forj=1:
ifdm4(j,i)==k&
dm4(j,i)~=0
a1(k)=a1(k)+1;
ifdm5(j,i)==k&
dm5(j,i)~=0
a2(k)=a2(k)+1;
a5=a1;
a6=a2;
a3=zeros(5,1);
a4=zeros(5,1);
j=1;
whilej<
6
a11=0;
a11=max(a1);
fori=1:
ifa1(i)==a11
a3(j)=a11;
j=j+1;
ifj>
5
break;
a1(i)=0;
a22=0;
a22=max(a2);
ifa2(i)==a22
a4(j)=a22;
a2(i)=0;
%%%%%%%%%%%%%%%%构造h1h2矩阵
h1=zeros(200,14);
h2=zeros(1,200);
y1=0;
y2=0;
y3=0;
y4=0;
forj=1:
x=0;
h1(i,1:
13)=dm3(j,1:
13)==dm3(i,1:
h1(i,14)=dm3(j,14)==dm3(i,14);
ifsum(h1(i,1:
13))>
h1(i,14)==1
x=x+1;
h2(j)=x;
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%%%%%%%%%%%%%%筛选h2矩阵得到h3h4矩阵
h3=zeros(1,200);
h4=zeros(1,200);
ifz14(i)==1
h3(i)=h2(i);
elseh4(i)=h2(i);
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
h5=h3;
h6=h4;
%%%%%%%%%%%%%%找六条重要样本%%%%%%%%%
y1=max(h5);
ifh5(i)==y1
h5(i)=0;
t1=i;
y2=max(h5);
ifh5(i)==y2
t2=i;
y3=max(h5);
ifh5(i)==y3
t3=i;
y4=max(h6);
ifh6(i)==y4
h6(i)=0;
t4=i;
y5=max(h6);
ifh6(i)==y5
t5=i;
y6=max(h6);
ifh6(i)==y6
t6=i;
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%找特殊样本结束
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%找特殊属性值
a3=zeros(6,1);
a4=zeros(6,1);
7
b1=zeros(6,1);
b2=zeros(6,1);
ifa5(i)==a3(j)
b1(j)=i;
ifa6(i)==a4(j)
b2(j)=i;
b3(1:
5,1)=b1(2:
6,1);
b4(1:
5,1)=b2(1:
5,1);
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%寻找结束
%%%%%%%%%%%%%%%%%%%%%验证程序开始
h7=zeros(103,14);
h8=zeros(103,14);
tc=0;
103
h7(i,1:
13)=dm3(200+i,1:
%%%%%%%%%%%%%%%%%%%%%%%%%%特殊样本
nc=0;
mc=0;
ncc=0;
mcc=0;
h8(i,1:
13)=h7(i,1:
13)==dm3(t1,1:
ifsum(h8(i,1:
=9
nc=nc+0.3;
13)==dm3(t2,1:
13)==dm3(t3,1:
13)==dm3(t4,1:
mc=mc+0.3;
13)==dm3(t5,1:
13)==dm3(t6,1:
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%样本结束
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%特殊属性值
ncc=0;
ifh7(i,j)==b3
(1)
ncc=ncc+3;
elseifh7(i,j)==b3
(2)
elseifh7(i,j)==b3(3)
ncc=ncc+2;
elseifh7(i,j)==b3(4)
elseifh7(i,j)==b3(5)
elseifh7(i,j)==b4
(1)
mcc=mcc+3;
elseifh7(i,j)==b4
(2)
elseifh7(i,j)==b4(3)
mcc=mcc+2;
elseifh7(i,j)==b4(4)
elseifh7(i,j)==b4(5)
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%属性值结束
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%挖掘算法
if(nc>
=0.6)
h7(i,14)=1;
elseif(mc>
h7(i,14)=2;
elseifnc>
=(mcc+2)
elseifmc>
nc
elseifncc>
mcc
elseifmcc>
ncc
else
%%%%%%%%%%%%%%%%%%%%%%%%%挖掘结束
%%%%%%%%%%%%%%%%%%计算正确率
h9=zeros(103,1);
h9=h7(:
14)==dm3(201:
303,14);
tc=sum(h9)/103
%%%%%%%%%%%%%%%%%%%%%%%算法完毕