From: Data reduction techniques for highly imbalanced medicare Big Data
Group a consists of: 1:81, 1:27, 1:2,500
Group b consists of: 1:9
Group c consists of: 1:3
Group d consists of: 1:1