MVAaerbh (MatLab R2007b)

Calculation of the AER (actual error rate) for price and clusters of Boston houses for the Boston housing data (bostonh.dat). Need function "hclust.m".

Download File

Zografia Anastasiadou, Awdesch Melzer

Applied Multivariate Statistical Analysis

hclust

Click the button to demonstrate a graph view. 
Notice: This content requires Java Runtime Environment.
Java Applet and JavaScript should be allowed on your browser.

Thu, April 19 2012 by Dedy Dwi Prastyo

hclust.m

bostonh.dat

- None

- The actual error rate for price and clusters of Boston houses.


clear all;
close all;
clc;
data = load('bostonh.dat');
xt = data;
xt(:,[1,3,5,6,8,9,10,14]) = log(xt(:,[1,3,5,6,8,9,10,14]));
xt(:,2) = data(:,2)./10;
xt(:,7) = (data(:,7).^(2.5))./10000;
xt(:,11) = exp(0.4*data(:,11))./1000;
xt(:,12) = data(:,12)./100;
xt(:,13) = sqrt(data(:,13));
xt(:,4) = [];
for i=1:13
    zz(:,i)=(xt(:,i)-mean(xt(:,i)))/(sqrt(var(xt(:,i))));
end
d = pdist(zz,'euclidean');
w = hclust(d); 
tree = cluster(w,'maxclust',2);
t1 = zz(find(tree(:,1)==1),:);
t2 = zz(find(tree(:,1)==2),:);
datat = [t1;t2];
n = length(tree);
  m1 = mean(t1);
    %mean of second cluster
  m2 = mean(t2);
    %mean of both clusters
  m = (m1+m2)/2;
miss1=0; %misclassified 1
miss2=0; %misclassified 2
corr1=0;
corr2=0;
i=0;
while i<length(t1)
    i=i+1;
    xg=datat((1:(length(t1)-1)),:) ;
    s=((length(t1)-1)*cov(t1)+(length(t2)-1)*cov(t2))/(length(zz)-2);
    alpha=inv(s)*(m1-m2)';
    if i<length(t1)
        if ((xg(i,:)-m)*alpha)<0
            miss1=miss1+1;
        end
    end
end
i=0;
while i<length(t1)
    xg=datat((1:(length(t1)+1)),:) ;
    s=((length(t1)-1)*cov(t1)+(length(t2)-1)*cov(t2))/(length(zz)-2);
    alpha=inv(s)*(m1-m2)';
    if i<length(t1)
        if ((xg(i+1,:)-m)*alpha)>0
            corr1=corr1+1;
        end
    end
    i=i+1;
end
i=(length(t1));
while i<length(data)
    i=i+1;
    xf=datat(((length(t1)):(length(data))),:) ;
    s=((length(t1)-1)*cov(t1)+(length(t2)-1)*cov(t2))/(length(zz)-2);
    alpha=inv(s)*(m1-m2)';
   if i<length(data)
        if (xf((i-length(t1)),:)*alpha)>0
            miss2=miss2+1;
        end
   end
end
i=(length(t1));
while i<length(data)
     i=i+1;
    xf=datat(((length(t1)+1):(length(data))),:) ;
    s=((length(t1))*cov(t1)+(length(t2))*cov(t2))/(length(zz)-2);
    alpha=inv(s)*(m1-m2)';
   if i<length(data)
        if (xf((i-(length(t1))),:)*alpha)<0
            corr2=corr2+1;
        end
   end
end
corr1
corr2
miss1
miss2
aer = (miss1+miss2)/length(zz) %AER (actual error rate)