Hi Seemant,
The persistence method combined with Fuzzy C-Means clustering can be used to select the best cluster.
I have written a dummy code to implement the same. The attached file named “bestCluster.m” contains the code.
In this code, I generated sample data using “randn” and performed Fuzzy C-Means clustering with “fcm”. The “U” matrix represents the fuzzy partition, with rows for data points and columns for clusters. The persistence vector is computed by summing squared membership values for each data point.
Then, I sorted the clusters based on the persistence values in descending order. The cluster with the highest persistence is considered the best. Data points belonging to the best cluster are extracted using a membership threshold of 0.5.
Finally, plotted the original data, best cluster, and cluster centers using “scatter” function.
This is just a dummy code, and you may need to adapt the code to your specific needs and adjust Fuzzy C-Means parameters and threshold as required.
You may review the official documentation related the functions used above:
I hope these steps will help you.
Thanks,
Ninad.
