New hard-thresholding rules based on data splitting in high-dimensional imbalanced classification

In binary classification, imbalance refers to situations in which one class is heavily under-represented. This issue due either a data collection process or because indeed rare population. Imbalanced classification frequently arises applications such as biology, medicine, engineering, and social sciences. this paper, for the first time, we theoretically study impact of sizes on linear discriminant analysis (LDA) high dimensions. We show that scarcity class, referred minority high-dimensionality feature space, LDA ignores yielding maximum misclassification rate. then propose new construction hard-thresholding rules based splitting technique reduces large difference between rates. proposed method asymptotically optimal. further two well-known sparse versions imbalanced cases. evaluate finite-sample performance different methods using simulations by analyzing real sets. The results our outperforms its competitors has comparable much smaller subset selected features, while being computationally more efficient.