Scatter-plot of data in which the cluster membership is coded by colors.

Hello, I have created a dendrogram of my given data.

NumCluster = 1566;
dist = pdist(alldata, 'euclidean');
GroupsMatrix = linkage(dist, 'complete');
clust = cluster(GroupsMatrix, 'maxclust', NumCluster);
E = evalclusters(alldata,clust,'CalinskiHarabasz')
[H,T,perm] = dendrogram(GroupsMatrix, 1566, 'colorthreshold', 'default');

I want to create now a scatter-plot of the data in which the cluster membership is coded by colors. I have tried to implement it like this

gscatter (alldata(:,1),alldata(:,2), E.OptimalY,'rbgk','xod')

Update:

The error is now gone but all the scatter plot has the same color. How can I cluster the membership by different colors? And is my E chosen correctly for the number of clusters? For my E I have 1566 cluster. I did not know if this is okay.

[the cyclist]

It can be difficult to diagnose issues without the data. Specifically, we don't see how E is defined. Can you upload the data, such that we can actually run your code and reproduce the error?

[Mark S]

Yes, I have attached the data

Answer 1

[KSSV]

gscatter (alldata(:,1),alldata(:,2),clust,'rbgk','xod')

Check the option E.OptimalY, it is empty []. So all the points are shown by same color/ maekers.

[Mark S]

Thank you. How can I define the number of clusters? I think I have actually to many (1566). How can I define the numbers of clusters?

[KSSV]

alldata = csvread('Data.csv') ; 
NumCluster = 10;  % <-----change cluster number here 

dist = pdist(alldata, 'euclidean');
GroupsMatrix = linkage(dist, 'complete');
clust = cluster(GroupsMatrix, 'maxclust', NumCluster); 

E = evalclusters(alldata,clust,'CalinskiHarabasz') ;
gscatter (alldata(:,1),alldata(:,2),clust)

[Mark S]

Thanks. It works fine. One additional question: How can I find an optimal cluster number? Is it best to vary the NumCluster myself or is there another method? I have found this in the matlab help:

https://it.mathworks.com/matlabcentral/answers/76879-determining-the-optimal-number-of-clusters-in-kmeans-technique

klist=2:500;%the number of clusters you want to try
myfunc = @(X,K)(kmeans(X, K));
eva = evalclusters(alldata,myfunc,'CalinskiHarabasz','klist',klist)
classes=kmeans(alldata,eva.OptimalK);

I get here for my optimalK=3. But I am not sure if this is ok. Is the calculation for the optimal cluster numbers so ok?