What to look for when preparing data for model training?

Question

Machine learning

Hello. I am new to machine learning. Therefore, a question arose. How to properly prepare data for training? I have a bunch of signs, but I don't know which ones are important and which aren't. I tried to calculate the correlation, but there all coefficients close to zero are obtained. Therefore, it is not clear what people are guided by when some variables are deleted, some are modified or aggregated, or even new variables are introduced. Most likely, knowledge of the subject area is required here. But what if they don't? Are there any signs other than correlation? Or is the only way to test it directly by simulating and seeing how it turned out? Thanks in advance!