A sequential three-way classification model based on risk preference and decision correction

The sequential three-way decision (S3WD) model, which merges three-way decisions and granular computing, is increasingly crucial in classification. The risk attitude to the decision process and result costs affects the decisive actions in the S3WD model. Furthermore, decision conflict arises when there is a discrepancy between coarse grained and fine-grained definite decision-making for the same object, which can potentially impact decision accuracy. However, current studies show incomplete risk preference research and a lack of decision correction strategies to address decision conflict. To address the limitation, four sequential three-way classifiers (S3WCs) are proposed. First, three prominent distance functions are employed to compute similarity classes for condition probability estimation. Second, optimistic, pessimistic, and weighted compromise sequential three-way classifiers are established to reflect the risk preference for the two types of costs. Third, four precision differences in the S3WCs are defined from local and global perspectives. An S3WC with decision correction is presented to improve precision by judging precision differences in adjacent granularity levels and the entire granular structure. Finally, a series of experiments are conducted to thoroughly analyze the characteristics and applications of these S3WCs. The superior classification performance of the proposed models on diverse datasets is demonstrated.

Automated summary

This paper investigates the application of the sequential three-way decision (S3WD) model in classification and proposes sequential three-way classifiers (S3WCs) to address risk preference and decision conflict. Experimental results demonstrate the superior classification performance of the proposed models on diverse datasets.