Geek Answers Handbook

SOM Clustering

I have data from 71 attributes and 17 instances. I want to classify them into six groups or classes. I tried with newsom( data, [ 6 6 ] ) .

The result is shown in the figures below. I can’t understand where the clusters are located and how can I find them programmatically?

enter image description hereenter image description here

I read all these documents regarding the SOM, but I never could understand how to get the clusters and data in these clusters? Therefore, please also indicate that when you answer my question.

+4
source share
2 answers

You need to carefully study the documentation about the structure of the return from the news (which are currently outdated) or the selforgmap function. In the IW field you can find the coordinates of the cluster N * N. For example:

 somnet = newsom( data, [ 6 6 ] ) my_clusters = somnet.IW;

myclusters will have N * N rows (in your case 6 * 6) and M columns are equal to the input sizes. It's all.

+1
source

Since you have a high instance ratio for matching nodes, and as a result, you have nodes that do not “win” the instance on the last card, you can split the map using these “empty” nodes. To learn more about clustering on a catfish, follow these steps: Clustering a self-organizing card

Keep in mind that SOM is not a random method of clustering, namely: you do not determine the number of clusters, the data will tell you about it.

+1
source

More articles:


All Articles