Question regarding SOM and unsupervised learning

I have got data from a survey that will be used to divide the participants into four groups. For this classification, should not I use unsupervised learning methods?

From this understanding, I tried with SOM. The result I got was astonishing: it gave raise to problem like the person with best result is falling into the worse group!

Another problem is that, I could not find the explanation of "SOM sample hits" cell numbering and their significances.