It is possible to find values and not labels in classification ?

Hi, I would like to know if it's possible to reverse a classification problem in Machine Learning. I have 3000 spectrums and 115 associated parameters. Usually, the method will be to find each of these parameters according the different measurements, but in my case I want to do the oppsosite and be able at the end to draw a spectrum in accordance with the input parameters.

I don't really know how to transform a 2-dimensions variable (freq,harm) into a specified label, I would be greateful if someone can help me on this problem. Thanks,

Jean Baptiste