Mathematical Modeling Series: Cluster Analysis

Cluster analysis

As the name suggests, "Like attracts like", is simply to have something similar personality polymerization classification. The samples were classified called Q cluster, referred to as the indicators classified R cluster analysis. More details refer to "modeling algorithms and Applied Mathematics" (Division keep Kui) this book.

Case Analysis

Q Cluster Analysis (From the book)

As salesman \ (w_i (i = 1,2,3,4,5) \) sales for the \ ((V_} {I1, I2} {V_) \) . To measure the distance between the point and the point of the absolute value distance, using the distance between the shortest distance measured class and class-based method, i.e.:
\ [D (W_i, w_j) = \ sum_ ^ {K} = {2}. 1 | v_ {ik} -v_ {jk
} |, D (G_p, G_q) = \ min \ limits_ {w_i \ in G_p, w_j \ in G_q} \ {d (w_i, w_j) \} \] is calculated by the formula distance distance matrix:

  1. All the elements of a class as \ (H_1 = \\ {W_1, w_2, w_3, W_4, w_5 \\} \) . Platform height of each class is 0, namely: \ (F (W_i) = 0, I = 1,2,3,4,5 \) . This time \ (D (G_p, G_q) = d (w_p, w_q) \)

  2. Platform height were taken 1,2,3,4 get a different classification, clustering map drawing:

matlab solving the code:

%代码1
clc,clear
a=[1,0;1,1;3,2;4,3;2,5];
[m,n]=size(a);
d=zeros(m);
d=mandist(a');  % mandist:求矩阵列向量组之间的两两绝对值距离
d=tril(d);  % 截取下三角元素
nd=nonzeros(d); %去除d中的零元素,非零元素按列排序
nd=union([],nd)  % 去掉重复的非零元素
for i = 1:m-1
    nd_min=min(nd);
    [row,col]=find(d==nd_min);
    tm=union(row,col); %row和col归为一类
    tm=reshape(tm,1,length(tm));%将数组tm变为行向量
    fprintf('第%d次合成,平台高度为%d时的分类结果为:%s\n',i,nd_min,int2str(tm));
    nd(nd==nd_min)=[];%删除已经归类的元素
    if length(nd)==0
        break
    end
end

operation result:

nd =

     1
     2
     3
     4
     5
     6
     
第1次合成,平台高度为1时的分类结果为:1  2
第2次合成,平台高度为2时的分类结果为:3  4
第3次合成,平台高度为3时的分类结果为:2  3
第4次合成,平台高度为4时的分类结果为:1  3  4  5
%代码2
clc,clear
a=[1,0;1,1;3,2;4,3;2,5];
y=pdist(a,'cityblock'); % 求a得两两行向量之间得绝对距离
yc=squareform(y)  %变换成距离方阵
z=linkage(y)  %产生等级聚类树
dendrogram(z)  %画聚类图
T=cluster(z,'maxclust',3) %把对象划分成3类
for i=1:3
    tm=find(T==i); %求第i类的对象
    tm=reshape(tm,1,length(tm)); %变成行向量
    fprintf('第%d类的有%s\n',i,int2str(tm)); %显示分类结果
end

operation result:

yc =

     0     1     4     6     6
     1     0     3     5     5
     4     3     0     2     4
     6     5     2     0     4
     6     5     4     4     0
z =

     1     2     1
     3     4     2
     6     7     3
     5     8     4
T =

     1
     1
     2
     2
     3

第1类的有1  2
第2类的有3  4
第3类的有5

R cluster analysis

Normal development of higher education in China Regions analysis

x1:每百万人口高等院校数
x2:每10万人口高等院校毕业生数
x3:每10万人口高等院校招生数
x4:每10万人口高等院校在校生数
x5:每10万人口高等院校教职工数
x6:每10万人口高等院校专职教师数
x7:高级职称占专职教师的比例
x8:平均每所高等院校的在校生数
x9:国家财政预算内普通高教经费占国内生产总值的比例
x10:生均教育经费

Matlab program to solve

clc,clear
data=load('raw_data.txt');  % 加载原始数据
b=zscore(data);  %数据标准化
r=corrcoef(b); %计算相关系数矩阵
d=pdist(b','correlation'); %计算相关系数d导出的距离
z=linkage(d,'average')  %按类平均法聚类
h=dendrogram(z);  %画聚类图
set(h,'Color','k','LineWidth',1.3) %把聚类图的颜色改成黑色,线宽加粗
T=cluster(z,'maxclust',6)  %把变量划分成6类
for i=1:6
    tm=find(T==i);  %求第i类的对象
    tm=reshape(tm,1,length(tm)); %变成行向量
    fprintf('第%d类的有:%s\n',i,int2str(tm));  %显示分类结果
end

operation result

z =

    3.0000    4.0000    0.0013
    5.0000    6.0000    0.0014
    2.0000   11.0000    0.0054
   12.0000   13.0000    0.0197
    1.0000   14.0000    0.0381
    9.0000   15.0000    0.1582
    7.0000    8.0000    0.2211
   10.0000   16.0000    0.3685
   17.0000   18.0000    0.6501


T =

     1
     2
     2
     2
     2
     2
     4
     5
     3
     6

第1类的有:1
第2类的有:2  3  4  5  6
第3类的有:9
第4类的有:7
第5类的有:8
第6类的有:10

Conclusions can be drawn from: x2、x3、x4、x5、x6the five indicators have greater relevance, is the first to come together so that the selected six indicators from the 10 indicators analyzed, and then these 30 areas Q and then through the six indicators cluster analysis.

Matalb solving the code

clc,clear
load raw_data.txt
raw_data(:,[3,6])=[ ];
raw_data=zscore(raw_data);
y=pdist(raw_data);
z=linkage(y,'average');
h=dendrogram(z);
set(h,'Color','k','LineWidth',1.3)
for k=3:5
    fprintf('划分成%d类的结果如下:\n',k)
    T=cluster(z,'maxclust',k);
    for i=1:k
        tm=find(T==i);
        tm=reshape(tm,1,length(tm));
        fprintf('第%d类的有%s\n',i,int2str(tm));
    end
    if k==5
        break
    end
    fprintf('=======================\n');
end

operation result:

划分成3类的结果如下:
第1类的有2  3
第2类的有4   5   6   7   8   9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30
第3类的有1
=======================
划分成4类的结果如下:
第1类的有4   5   6   7   8   9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  26  27  28  29
第2类的有25  30
第3类的有2  3
第4类的有1
=======================
划分成5类的结果如下:
第1类的有25
第2类的有30
第3类的有4   5   6   7   8   9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  26  27  28  29
第4类的有2  3
第5类的有1

Guess you like

Origin www.cnblogs.com/CH42e/p/11961432.html