In machine learning, how do you know when you will need to scale data? I am not asking if I should scale data! I am asking at what range differences should I scale data? I have not found anything that addresses this second question.
I have a data set that has mostly categorical features with values 0, 1, 2. Some categorical features range from 1-6. In addition, there are 3 numerical variables ranging 0-8, 0-13, and 0-1000. When I scale this 3rd numerical variable (0-1000), I get the range to be 0-46. Will the max of 46 still be too large considering that the max for most variables is less than 10?