When working on a problem at what stage do you think is best to check/validate assumption? Is it during EDA, data pre-processing, before building the training pipeline or later?
Typically you understand the data, clean it (i.e. EDA). Its after this that you decide what models to experiment with. If linear reg is the model, its good check some of the assumption to see how things would play out.
When working on a problem at what stage do you think is best to check/validate assumption? Is it during EDA, data pre-processing, before building the training pipeline or later?
Typically you understand the data, clean it (i.e. EDA). Its after this that you decide what models to experiment with. If linear reg is the model, its good check some of the assumption to see how things would play out.