Super-Agers vs Cognitive Decliners

Abstract

Classification models rely on the class labels to learn and estimate the boundaries between different classes. Sometimes the labels are not available, or the labeling procedure is complicated. In this study, we proposed a hybrid machine learning-optimization framework to optimally assign the class labels to the observations. The proposed algorithm incorporates the information from a separate dataset called shadow dataset to find the class labels. The algorithm utilizes Bayesian Optimization to optimize the decision variables pertinent to the labeling procedure.