Is there any criteria or algorithm by which two different features (in specific high and low channel gain)users needs to be grouped together.

I am working on NOMA system and I want to group the users into clusters.

I tried with PSO based K-means clustering which works on minimizing the distance and thereby grouping two similar features of users which means both users have low channel gain.

I would like to check is there any algorithm or method that groups two different features of users .