machine learning - How to choose the right normalization method for the right dataset? -

There are several common ways to choose from L1 / L2 ideal, jade-score, minimum-maximum what information How can one choose the appropriate general method for a dataset?

I did not pay much attention to generalization before, but I got a small project where its performance is huge, not affected by the parameters or options of ML algorithm, but by the way I am normalized data I like to wonder but in practice it can be a common problem. So can someone give good advice? Thanks a lot!