Imbalanced data is when classes do not have similar proportions and this can occur in multiple common problems like predicting mortality where the vast majority of data points will have “Survived” instead of “Died” as the target variable, so imbalanced data occurs and trained models will probably have a bias towards the majority class.
Oversampling the minority class is a possible solution to have a more balanced dataset. However, splitting the dataset into train/validation/test must be done before applying oversampling to avoid data leakage.
check this article that discusses in depth 5 oversampling techniques.