I am using hierarchical clustering algorithm to cluster my dataset with different number of clusters. For instance,
a= [1;2;3;4;5;20;21;22;28;29]
Z=linkage(a,'ward')
[clusterIndexes]=cluster(Z,'maxclust',2)
this snippets clusters data into two at where first cluster holds 1,2,3,4,5. Lets call this cluster A and the second one holds 20,21,22,28,29 which is cluster B.
When I run the following script and cluster data into 3
a= [1;2;3;4;5;20;21;22;28;29]
Z=linkage(a,'ward')
[clusterIndexes]=cluster(Z,'maxclust',3)
It gives me clusters of (1 2 3 4 5)= cluster X,(20,21,22) = cluster Y,(28,29) = cluster Z.
How can I demonstrate programmatically that cluster B splitted into Cluster Y and Cluster Z?
Sorry for the naive question I am very new to matlab.
You can use setxor to identify any differences between the two clusters. User union to merge clusterY and clusterZ when comparing to clusterB. Since the result is empty the two clusters contain the same set of numbers. If there were any differences between the two it would be output by setxor.
clusterB = [20 21 22 28 29];
clusterY = [20 21 22];
clusterZ = [28 29];
setxor(clusterB, union(clusterY, clusterZ))
ans =
1×0 empty double row vector
Suppose for example clusterB had an additional number, you can see the results below.
clusterB = [5 20 21 22 28 29];
setxor(clusterB, union(clusterY, [clusterZ]))
ans =
5