Latent Representation for Silhouette Score to Determine Number of Clusters

[Radwaadel267]

Thank you for the great package and method.

I was wondering if I do not have a pre-determined number of clusters, when using silhouette score, should I use the latent representation or original data to determine the optimal number of clusters. Thank you!

[longyahui]

Hi,

Thank you for the question.

We can determine the number of clusters according to unsupervised metric 'Silhouette score'. In general, the higher Silhouette score means better clustering results. Hope it helps. Thank you again.

Best regards,
Yahui

[Radwaadel267]

Thank you for your answer, but I was wondering which data do you use to measure it, is it original data or transformed data.

[longyahui]

Welcome. We can use the latent representation, output of the GraphST model. Thanks.

[Radwaadel267]

Thank you so much for your quick reply!

[longyahui]

Welcome. I am very more than happy to help if you have further questions. Thanks.

[Radwaadel267]

Thank you!