Sample size distribution of positive (persistent opioid use) and negative classes (no persistent opioid use) in the original training set (80% of total sample) vs SMOTE dataset
SMOTE, Synthetic Minority Oversampling Technique.