Estimating the minority class proportion with the ROC curve using Military Personality Inventory data of the ROK Armed Forces
Meesun Sun, Kwang‐Hyun Choi, Sungzoon Cho · Journal of Applied Statistics · 2015
The Republic of Korea Armed Forces includes maladjusted conscripts such as the mentally ill, the suicidal, the imprisoned, and those determined by the military commander to be maladjusted. To counteract these problems, it is necessary to identify the maladjusted conscripts to determine who among them would qualify for exemption from active military service or need special attention. We use the Military Personality Inventory (MPI) to make this prediction. Such a prediction presents a kind of class imbalance and class overlap problem, where the majority fulfil active service and the minority are maladjusted, the latter being discharged early from active service. Therefore, most classification algorithms are likely to show low classification performance. As an alternative, this study demonstrates the effective utilization of the receiver operating characteristics curve using MPI data to estimate the maladjusted proportion of persons sharing similar MPI test results. We confirm that the suggested method performs well using the real-world MPI data set. The suggested method is very useful to estimate the proportion of conscripts maladjusted to military life and can help in the management of such persons subject to conscription.