Incorporating logistic regression to decision-theoretic rough sets for classifications

Text of abstract Logistic regression analysis is an effective approach to the classification problem. However, it may lead to high misclassification rate in real decision procedures. Decision-Theoretic Rough Sets (DTRS) employs a three-way decision to avoid most direct misclassification. We integrate logistic regression and DTRS to provide a new classification approach. On one hand, DTRS is utilized to systematically calculate the corresponding thresholds with Bayesian decision procedure. On the other hand, logistic regression is employed to compute the conditional probability of the three-way decision. The empirical studies of corporate failure prediction and high school program choices’ prediction validate the rationality and effectiveness of the proposed approach.

► We integrate logistic regression and DTRS to provide a new classification approach.
► We propose two novel integrated classification models to solve the binary misclassification problem and multiple classification problem.
► We provide a human-machine viewpoint in three-way decisions.