The research on the classification of the incomplete information system

An approach to solve the classification problems of the incomplete information system, which is in accord with the human cognitive customs, is proposed in this paper. This approach can decompose the system into two parts - attributes complete and incomplete systems. A classifier for the complete system is obtained by learning the attributes complete samples. As for the attributes incomplete information system, the projecting is used to obtain a new decision system, which is processed into a decision consistency system by rough granular calculating. The relearning of this decision consistency system helps to form a new classifier. In the process of recognition, a corresponded classifier is chosen according to the test sample. The adoption of this approach largely expands the extent of various classification algorithms´ applications that are not directly used for incomplete samples classification and discover some knowledge of the incomplete information system. The experimental results prove the effectiveness of the approach.