Superiority of three-way decisions from the perspective of probability

Three-way decisions (3WDs) is a typical method for dealing with uncertain issues. It is essentially an extension of two-way decisions (2WDs). What is the superiority of 3WDs/S3WDs over traditional 2WDs? Only a few studies have analyzed the theoretical superiority of 3WDs over traditional 2WDs. The motivation of this paper is to theoretically analyze the superiority of 3WDs over 2WDs in dealing with classification problems. From the perspective of probability, 3WDs is compared with 2WDs for a thorough discussion and analysis of the superiority of 3WDs in this paper. First, it is proved that increasing information can effectively improve classification precision in terms of the sample mean, misclassification probability, and interval of uncertain classification. Second, a novel 3WDs model/sequential three-way decisions model (3WDM-CLSM/S3WDM-CLSM) based on confidence level and sample mean is proposed, and the corresponding method for calculating the pair of thresholds based on the confidence level is presented. Third, the superiority of 3WDs compared to 2WDs is analyzed in terms of classification precision and cost of information acquisition (CIA). Finally, experiments are completed to show that the 3WDs model can effectively reduce the CIA, and its classification accuracy is close to that of the 2WDs model.

Automated summary

This paper analyzes the superiority of three-way decisions (3WDs) over traditional two-way decisions (2WDs) and proposes a 3WDs model based on confidence level and sample mean. Experiments show that this model can effectively reduce the cost of information acquisition and has similar classification accuracy to the 2WDs model.