I am unable to run random forest classification using MTD_MSIL1C data . How to atmospherically correct the data

sen2cor converts L1B products into atmospherically corrected L2A products. In many cases, L2A can be already downloaded.

But the more important step is to resample to a common resolution first using the S2 Resampling operator and then reproject (without tie-point grids) to WGS84: