There is an interesting article http://www.mql5.com/en/articles/163 Has anyone tried this algo for trading pupose?It says ''GNG is superior to "K-means" clustering''
NN's are mostly overkill. Trading is playing human psychology. S/R is a very simple concept that works quite well. I'm sure there are NN's that can determine profitable S/R levels but I've yet to see a public NN that doesnt overcomplicate basic trading logic.