The movement strategy of three-way decisions based on clustering

Abstract

The movement strategy is a crucial issue of three-way decisions, which transfers objects in the unfavorable region to the favorable region. For object-based movement strategy, each object in the unfavorable region has one particular movement rule, so many objects can be successfully moved but the movement process is complex. Contrarily, for region-based movement strategy, all objects in the unfavorable region use the same movement rule, so the movement process is simple but few objects can be successfully moved. Therefore, to design a movement strategy with lower complexity and higher success movement rate, we propose a movement strategy based on clustering. First, we divide objects in the unfavorable region into several smaller clusters. Second, for each cluster, to define a reference object as its representative, we design four criteria: the highest frequency of global attribute value, the highest frequency of local attribute value, the object with maximum distance, the object with minimum distance. Third, a movement rule is developed for each cluster and the movement rules of all clusters form cluster-based movement strategy. Finally, to evaluate the effectiveness of the proposed movement strategy, we define a utility function. Experimental results demonstrate the effectiveness of the proposed methods.