Research Of Incomplete Decision System Data Mining Base On Granular Computing

With the development of information technology and Internet, the huge database increases day by day. In order to obtain the valuable information and knowledge in the huge database, Data Mining (DM) emerges as the times require. Accordingly, the DM technology, which is widely used, has become the hotspot of research both at home and abroad. Recently, some mature methods of DM only deal with the complete information. During the practical DM, the data dealt with is always incomplete to some extent. If the incomplete information is dealt with directly by the DM method for the complete information system, it will end with unreasonable result, even wrong information and knowledge. Therefore, the researches on DM based on incomplete system have certain practical significance.In recent years, many domestic and international scholars have researched into the application of Granular Computing (GrC) in DM, and also put forward some models and methods. Then, GrC is widely used in DM. The method of GrC is applied in DM of incomplete decision system. Attribute reduction method and classification analysis method of incomplete decision system under granular expression are put forword in this paper, which has validated by experiments. The major work is as follows:1. With the Application of the basic theory of RoughSet and GrC, we researches on the method of attribute reduction, which is based on the model of GrC under tolerance relations. The determinant conditions of attribute necessity and importance under granular expression are put forward and an attribute reduction algorithm under granular expression is proviedes.2. The method of Decision-tree Classification is put forward in this paper. The choice criterion of the splitting attribute of Decision-tree under granular expression and the condition for stopping Decision-tree growing, are put forward. We provides a kind of Granular Computing Decision-tree Classification (GrCDC) algorithm. Finally, take experiments to prove the validity of the GrCDC algorithm.