Ugh, I feel you. Trial and error sucks.
One hack: Use feature importance scores from a quick model (like XGBoost) to rank datasets. If the model can’t learn anything, maybe ditch that dataset.
Also, DataWrangler is a cool tool for cleaning and prioritizing. Not perfect, but helps.
One hack: Use feature importance scores from a quick model (like XGBoost) to rank datasets. If the model can’t learn anything, maybe ditch that dataset.
Also, DataWrangler is a cool tool for cleaning and prioritizing. Not perfect, but helps.
