The Research Of Rough Set In Discriminant Rule Mining

Data mining, which is the combination of database technique, artificial intelligence and machine learning, is used to discover implicit, previously unknown, potentially useful knowledge from a large amount of data accumulated in database, is a promising research area. From the appearance of data mining, a lot of research has been made to it in native and outside, but most of algorithms heavily depend upon the quality of data resource, in order to overcome this weakness, the dissertation studies rough set model of information system, explores new knowledge discovery approaches hiding in the data which have above properties.Firstly, the dissertation deeply studies the rough set model, and indicates low approximate set of a concept including sufficient condition which can deduce this concept, data in this set can be generalized to discriminant rule.Secondly, the dissertation studies the technique of mining discriminant rule in rough set model of information system. In the first place, on the base of studies concept ascending and attribute dependency, we forward a method to select attribute value partition, which can effectively resolve the problem of minimum faith and minimum support, in order to induct potential knowledge from standard discrete symbolic data, the dissertation studies join operation for generating candidate set and delete operation for deleting redundant equivalence class, and propose an recursion algorithm to realize highly generalization of data. Lastly, using the property of adjacent numeric interval can be united into one interval, we put forward a greedy algorithm of rule reduction.Finally, we design and implement a data mining system, which can defines task of mining according to the question of user and find effective discriminant rule from database.