I am working on a Machine Learning project applied to a financial time series.
Initially, I grabbed features (open, high, low, close) and implemented a Random Forest.
One of the subsequent tasks says to explore if the close-price serie is non-stationary, and if its not, to differentiate it turning it into stationary. I was able to differentiate it fine (ran the Augmented dickey fuller test to verify as well). But I have the following question: When differentiating a time-serie and working with machine learning, are you afterwards supposed to train the model again using this new differentiated serie? I'm having problems to understand the practical application of differentiating the time-serie (i.e., turning the serie into stationary)
