Neural network, how to perform normalization when the answer is unknown?

Question

Neural networks

Tell me if I want to receive some specific answer from the neural network, and not YES (1) or NO (0), let's say I'm trying to predict the movements of the exchange rate, I take the last 20 days where the price was from 60 to 65 and normalize in the interval from 0 to 1, i.e. 60 is the minimum, 65 is the maximum, but if the next day can be 66 or even 68, it will also go beyond the normalized framework, what to do in such cases? Is it worth adding some values ​​(reserves) to the minimum and maximum values ​​as a percentage, i.e. let's say the minimum value is minus 20% and the maximum value is +20%

minValue = 60 - (((65-60) / 100) * 20);
maxValue = 65 + (((65-60) / 100) * 20);

You can only please in simple language, on the fingers. Thanks a lot.

Answer 1

https://en.wikipedia.org/wiki/Training,_test,_and_...