Three sequential multi-class three-way decision models

Abstract Decision conflict is a crucial issue of three-way decisions of multi-class. To overcome this limitation, there are mainly two ways. One is conflict resolution after decision-making, and the other is conflict resolution before decision-making. Considering insufficient information is a main cause of decision conflict, we propose a sequential approach to solve this problem by adding more detailed information step by step, which is conflict resolution during decision-making. Combining sequential approach with three methods to handle multi-class decision, which are converting m-class decision to m two-class decisions, based on pair-wise comparisons of m decision classes and directly making three-way decisions for each decision class of m-class simultaneously, we propose three sequential multi-class three-way decision models. At each level of three models, we define conflict region. In the sequential process of decision-making, for each model, only the objects belonging to one positive region with certain decision are assigned into corresponding decision classes, the objects in conflict region will be processed at the next level by adding more information sequentially. Finally, we compare the performance of three conflict resolutions and the performance of the proposed three models. Experimental results validate the effectiveness of the proposed three sequential multi-class three-way decision models.