Clusters in HitMapView

[anjomshoaa]

Hi,

I have a question about the HitMapView. I do clustering as follows and the output is perfectly fine:

cl = som.cluster(n_clusters=3)

But the HitMapView always uses the default number of clusters which is 8 and I could not find a way to change it:

h = hitmap.HitMapView(10, 10, 'hitmap', text_size=8, show_text=True)

How can I specify number of clusters in a HitMapView to a different number?

Regards
Amin

[sevamoo]

Hi @anjomshoaa there was an issue in som.cluster() that it didn't set the cluster labels. Now should be fine.
Could you please check it out?

[anjomshoaa]

Thanks Vahid jan,

Yes, it is working perfectly now.

Best regards
Amin

[gsvijayraajaa]

Is there a way to get the index of the training data for the individual cluster labels?

[sevamoo]

by clustering you get the labels of nodes. By project function you get the bmus. So, then use something like this
cl = som.cluster(n_clusters=3)
bmus = som.project()
cl_data = cl[bmus]

[gsvijayraajaa]

Thanks for the direction. Is there a som.project() function available? Regards, Vijay Raajaa GS

…

On Tue, Aug 29, 2017 at 7:54 PM, Vahid Moosavi ***@***.***> wrote: by clustering you get the labels of nodes. By project function you get the bmus. So, then use something like this cl = som.cluster(n_clusters=3) bmus = som.project() cl_data = cl[bmus] — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#24 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AEIJ1hcZKAdS6-VjaXFtqoqibRXxE9yqks5sdB8zgaJpZM4JCl4u> .

[sevamoo]

Yes, there is one.

SOMPY/sompy/sompy.py

Line 433 in c673b19

def project_data(self, data):

[gsvijayraajaa]

Using the project_data function , I need to project back the training data to the trained model? Also, from the bmus, how do I tie it back to the original training data index? Essentially I would like to find the list of training data index corresponding to the cluster label.

[vishvanath45]

@gsvijayraajaa Did you get a way get the output category label back to data.

[vedavikas06]

to cluster the neurons into 4 clusters

cl = som.cluster(n_clusters=4)

this will map the data X to the respective neurons in som

bmus = som.project_data(X)
print(bmus)

this gives the corresponding cluster number for the data in X

cl_data = cl[bmus]
print(cl_data)

[vishvanath45]

Thanks it worked!

[businessglitch]

Hey @sevamoo, really need help with this issue. I have a dataset of length 50,000 when i do cl = som.cluster(n_clusters=2)

cl returns lables which is fine but the length of 1120. Which is understandable because the data points are under a cluster.

but when I do cl_data = cl[bmus], this creates duplicate indexs like there would be 7 1's 5 2's and so on, which is not correct because the model wouldn't use duplicated values. Any advice on this would be appreciated. I believe it has something to do with the length of cl the 1120

[nehamalik1978]

I have used the below code: to find label and index of cluster:
cl = som.cluster(n_clusters=3)
bmus = som.project()
cl_data = cl[bmus]

How do I find which cluster my data belongs to?
I tried:

N_CLUSTERS = 3
clusters = [input_data[km == i] for i in range(N_CLUSTERS)]
for i, c in enumerate(clusters)
print('Cluster {} has {} members: {}...'.format(i, len(c), c[0]))

but it is showing error: item length not match

[sevamoo]

I believe this is the answer you are looking for: cl_data = cl[bmus]

[kedarnaik33]

Hi Sevamoo,

I have a question, in the som.cluster(n_clusters = 8), I have set 8 clusters and it gives me clusters from 0 to 7.

However, my test set labels are from 1 to 8 and when I use accuracy_score it obviously gives me a wrong accuracy.

I wanted to know if there is a way to set the cluster number from 1 to 8 ?

Would really appreciate your help.

Sincerely,
Kedar

[akol67]

Using the project_data function , I need to project back the training data to the trained model? Also, from the bmus, how do I tie it back to the original training data index? Essentially I would like to find the list of training data index corresponding to the cluster label.

Did you have some help to accomplish that list of training index?