Mean Normalization VS other forms of Feature Scaling

I’m a bit confused on which specific cases Mean Normalization could be best for. As in, when is it best practice to use Mean Normalization over other forms of scaling?

I think Min-Max normalization work best when our observations and data are not presented within a Normal/Gaussian distribution. Meaning that the values are not centered around a mean with a unit standard deviation from that said mean. With the same train of thought, I think that Z-score normalization/Standardization works better when dealing with data in a Gaussian distribution–logical enough, I guess. Yet, I can’t think of any case where I would use mean normalization over other forms of scaling, probably because I’m a newbie in the Stats.

Any help would be appreciated. Thanks!

Min/max vs. standard deviation comes down to how much you want to emphasize the impact of outliers in the data set.

Min/max will include the full range of any outliers. Standard deviation will reduce the impact of outliers.

Some domain knowledge might be necessary to support the decision for any specific data set. Or you can try both and see which one gives better results.

Oh yeah, I didn’t take that into account. Thanks for the clarification.