I took a sample of my training data and balanced it and then trained my model. The results obtained are more accurate than using the whole set of train data (balanced or imbalanced). My question is: what could explain this result? (the entire set of data does not contain any noise).
Thanks in advance.