Weighting Clustering Data using Self Organizing Maps

[nn_4]

Hi, I am doing clustering using Self Organizing Maps in Knime. I am looking to weigh various input variables so I get clusters that reflect strategic importance. Can anybody help me with that?

[lisovyi]

Hi,

are you using this node?
https://kni.me/n/Ztd8_jiVdtZ2roU8
If yes, it seems that id does not allow to define weights for variables. However, what you can do is to implement custom preprocessing of the input data in a way to first standardize all feature and then to scale them with the weights that come from your domain knowledge. Afterwards you will want to disable normalisation of the input features in the SOM node itself, of course.

Alternatively, you can try out implementations in python or R, as knime provides a simple intergation with both. However, i do not know if weighting would be directly available in an implementation in any of those

Best,
Mischa

[nn_4]

Hi,

I am. Thanks for your suggestion. I think I will test with assigning weights to the input data based on my importance for the input factors. And then see if the output develops as I was hoping for it to develop. If it doesnt work the way I would expect it to then I will try Python or R.

Thanks
Namra.