Studies on Some Details of Algorithm IGRS

In containing order rough set methodology (CORS), ordered attribute `criterion´ is introduced. Some terminologies on rules or rules set, such as robust, minimality, completeness, mutuality degree, and conflict are discussed. The rules generation algorithm IGRs is given and the details of algorithm IGRs are studied. Heuristic knowledge, which is mutuality degree of a condition item with a decision part, is used to choose condition item when generating rules. In primary and modified IGRs, two kinds of mutuality degree, simple and weighted mutuality are introduced respectively. In addition, the variable precision method is used to solve the conflict problem in modified IGRs. By experiments, the effects of two kinds heuristic knowledge and different weight values in synthetic mutuality on algorithms properties are shown, such as time consumption, calculation precision etc. The performances of IGRs with the primary and new conflict solution are compared by experiments. The conclusion is that the weighted mutuality degree is more sound and the choice of appropriate weight values in it are important to optimize the quality of rules set. The variable precision method for dealing with conflict when generating rules is more reasonable. Both two modifications to primary IGRs make the performance of IGRs enhanced and the quality of rules set better. Algorithm IGRs still need further improvement