For example let take 'home_rentals' dataset and let train predictor to predict 'rental_price'. I made test of three cases:
train on data 'as is', without any changes. I got right results.
train on data with serial column with name not equal to 'id'. In this case results good too.
train on data with serial column with name 'id'. In this case results absolutely bad. 'rental_price' in prediction results in most cases is negative. min/max values looks like random values.
For example let take 'home_rentals' dataset and let train predictor to predict 'rental_price'. I made test of three cases: