Mashinali o‘qitishga kirish Nosirov Xabibullo xikmatullo o‘gli Falsafa doktori (PhD), tret kafedrasi mudiri

xlabel('Sepal Length'); ylabel('Sepal Width'); zlabel('Petal Length'); view(-137,10); grid on sidx = grp2idx(species); miss = find(cidxCos ~= sidx); plot3(meas(miss,1),meas(miss,2),meas(miss,3),'k*'); legend({'setosa','versicolor','virginica'}); hold off

It's clear from this plot that specimens from each of the three clusters have distinctly different relative sizes of petals and sepals on average. The first cluster has petals that are strictly smaller than their sepals. The second two clusters' petals and sepals overlap in size, however, those from the third cluster overlap more than the second. You can also see that the second and third clusters include some specimens which are very similar to each other.

Clustering Fisher's Iris Data Using Hierarchical Clustering

K-Means clustering produced a single partition of the iris data, but you might also want to investigate different scales of grouping in your data. Hierarchical clustering lets you do just that, by creating a hierarchical tree of clusters.

First, create a cluster tree using distances between observations in the iris data. Begin by using Euclidean distance.

eucD = pdist(meas,'euclidean'); clustTreeEuc = linkage(eucD,'average');

The cophenetic correlation is one way to verify that the cluster tree is consistent with the original distances. Large values indicate that the tree fits the distances well, in the sense that pairwise linkages between observations correlate with their actual pairwise distances. This tree seems to be a fairly good fit to the distances.

cophenet(clustTreeEuc,eucD)

Clustering Fisher's Iris Data Using Hierarchical Clustering

To visualize the hierarchy of clusters, you can plot a dendrogram.

[h,nodes] = dendrogram(clustTreeEuc,0);

h_gca = gca;

h_gca.TickDir = 'out';

h_gca.TickLength = [.002 0];

h_gca.XTickLabel = [];

The root node in this tree is much higher than the remaining nodes, confirming what you saw from K-Means clustering: there are two large, distinct groups of observations. Within each of those two groups, you can see that lower levels of groups emerge as you consider smaller and smaller scales in distance. There are many different levels of groups, of different sizes, and at different degrees of distinctness.