First you need to be sure that your media is not autofluorescent, as adviced in your previous thread (https://www.researchgate.net/post/How_to_minimize_autofluorescence_of_yeast_cells)
Then, I would use a non-GFP tagged strain to calculate the mean autofluorescence per size of the cell (find the relation between FSC-A and GFP-A for example, let's call it a=GFP-A/FSC-A).
Next step is to apply a corrective term on your GFP sample, substracting GFP-A with a*FSC-A.
But better find a way to improve your signal/autofluo ratio!
Basically you generate emission curves by using your spectral detector to sample the emission in the sample over a series of wavelength "bins". Then you create the curves by first making an ROI over the background area where there is only autofluorescence, and another area where there is GFP signal. The software will create an emission curve based on the intensity in each bin and these can be stored. Then, use these curves to separate the autofluorescence from the GFP signal. The exact process is software specific.