About coding for optimal no of cluster

hi all

i create 5 cluster(By Fuzzy C Means)

now i am moving for next step, my next step is choosing optimal no. of cluster.

i am choosing silhouette method for optimal no. of cluster

i want help about coding for silhouette method, if anyone known please help me

Thank you

Answer 1

[Paras Gupta]

Hi Seemant,

I understand that you want to find the optimal number of clusters using the Silhouette Method with Fuzzy c-means as the clustering function.

You can refer the code below to achieve the same in MATLAB.

% Generate dummy data
data = [randn(100,2)+2; randn(100,2)-2];

% Set the range of cluster numbers to evaluate
minClusters = 2;
maxClusters = 10;

% Initialize variables to store optimal number of clusters and maximum silhouette score
optimalNumClusters = 0;
maxSilhouetteScore = -Inf;

% Perform fuzzy c-means clustering for different numbers of clusters
for k = minClusters:maxClusters
    
    options = fcmOptions(NumClusters=2,Verbose=false);
    [centers, U] = fcm(data, options);
    
    % Calculate the fuzzy membership values
    [~, labels] = max(U);
    
    % Calculate the silhouette values for each data point
    silhouetteValues = silhouette(data, labels);
    
    % Calculate the average silhouette score
    avgSilhouetteScore = mean(silhouetteValues);
    
    % Update the optimal number of clusters if a higher silhouette score is found
    if avgSilhouetteScore > maxSilhouetteScore
        maxSilhouetteScore = avgSilhouetteScore;
        optimalNumClusters = k;
    end
end

% Display the optimal number of clusters
disp("Optimal number of clusters = " + optimalNumClusters);

Optimal number of clusters = 2

Please find below the documentation links of the functions used in the code:

• fcm - https://www.mathworks.com/help/fuzzy/fcm.html
• silhouette - https://www.mathworks.com/help/stats/silhouette.html

You can also refer to the following documentation on "Silhouette criterion clustering evaluation object" to use the silhouette method with other clustering functions.

https://www.mathworks.com/help/stats/clustering.evaluation.silhouetteevaluation.html

Hope this helps.