Abstract: By means of comparing the upper approximation sets, the lower approximation sets and accuracy measures of six covering-based rough set models, the six models were systematically studied. Three conclusions were obtained. First1y, among six covering upper approximation sets, the second one is the largest, and the first five upper approximation sets all have inclusion relationship, except the third and fourth ones. Secondly, there is inclusion relationship between two lower approximation sets. Thirdly, among the accuracy measures of six covering-based rough set models, the second one has the lowest accuracy measure, and the fifth one has the highest accuracy measure, but it has no relationship with the sixth one. I3y some illustrative examples, all the conclusions gain demonstrations. The comparative study on different rough set models provides a better understanding of these models and some references for model selection in different applications.

Key words: Covering, Rough set models, Accuracy measure, Model comparison