Algorithm Studies of Rules Generation in CORS

In containing order rough set methodology (CORS), ordered attribute `criterion´ is introduced. Criterion is related semantically with decision attributes, which results in producing more rational and significant rules utilizing dominance relations. In this paper, some terminologies and properties on rules or rules set, such as robust, minimal, complete are discussed, the state of art in algorithms on rules generation are analyzed, and algorithms GRs and IGRs are proposed. We compare these two algorithms by experiments in time complexity, rules count and accuracy. We find that GRs can generate all minimal rules, but time complexity is high and IGRs´ efficiency is better, but quality of rules are inferior to that of GRs. We also discuss the completeness of rules set and present a viewpoint that there are three hierarchies for completeness. In accuracy calculation, three formulas for accuracy calculation and two approaches for experimental test are given