Hi Dro, I am not entirely sure what you mean here but I think so long as you are using the data provided in the dataset then anything goes and creativity and exploration is encouraged.
My concern with this approach though is 1: what are you using the target to predict? Because the task is to predict this value and then 2: How do you cope with the test set when you are not given the target values.
Good luck with trying your approach and hope this helps
Hi Dro, I am not entirely sure what you mean here but I think so long as you are using the data provided in the dataset then anything goes and creativity and exploration is encouraged.
My concern with this approach though is 1: what are you using the target to predict? Because the task is to predict this value and then 2: How do you cope with the test set when you are not given the target values.
Good luck with trying your approach and hope this helps
For example, I can use the previous (or next) target value provided by train as a prediction. In real life most of such models will be useless