Blending Propensity Score Matching and Synthetic Minority Over-sampling Technique for Imbalanced Classification

Real world data sets often contain disproportionate sample sizes of observed groups making the task of prediction algorithms very difficult. One of the many ways to combat inherit bias from class imbalance data is to perform re-sampling. In this paper we discuss two popular re-sampling approaches proposed in literature, Synthetic Minority Over-sampling Technique (SMOTE) and Propensity Score Matching (PSM) as well as a novel approach referred to as Over-sampling Using Propensity Scores (OUPS). Using simulation we conduct experiments that result in statistical improvement in accuracy and sensitivity by using OUPS over both SMOTE and PSM