If you use a random state to split your test and train data then the model it leads to future look ahead bias because the model already knows what will happen. I havent yet, but if you try the model without randomizing the test set form the dataset it would preform significantly worse.
If you use a random state to split your test and train data then the model it leads to future look ahead bias because the model already knows what will happen. I havent yet, but if you try the model without randomizing the test set form the dataset it would preform significantly worse.