I want to make 4 clusters using k mean clustering and for each of the 4 clusters i want to set centroid initially and then accordingly make the cluster. how can I.

I want to make 4 clusters using k mean clustering and for each of the 4 clusters i want to set centroid initially and then accordingly make the cluster. how can I.

Answer 1

[Walter Roberson]

https://www.mathworks.com/help/stats/kmeans.html

numclusters = 4;
Initial = numeric matrix that is numclusters x number of columns in YourData
[idx, C] = kmeans(YourData, numclusters, 'Start', initial);

[Maimoona Asad]

What is initial..?

[Walter Roberson]

"for each of the 4 clusters i want to set centroid initially"

Those initial centroids are what will go into Initial.

[Sam Chak]

@Walter Roberson

Guess @Maimoona Asad wants to 'estimate' 4 centroids for the given set of data, probably because she does not have a clear idea how many clusters there should be.

The centroid estimations are intended to indicate the mean locations of the clusters, even though this initial guess for the cluster centers may be inaccurate.

With the initial cluster centers, then she can apply the k-means clustering technique as you suggested.

[Walter Roberson]

The 'Start' option has several possibilities to estimate cluster centers in different ways; https://www.mathworks.com/help/stats/kmeans.html#buefs04-Start

Answer 2

[Sam Chak]

Hi @Maimoona Asad

Perhaps you can try the pdist2 method outlined in this link to find the centroids. Hope it helps.

https://www.mathworks.com/help/stats/pdist2.html#mw_1d953621-cd51-4b4e-8462-a34db4901234

Answer 3

[Sam Chak]

@Maimoona Asad

Have you successfully estimated the centroids and closed this issue?