Research On Dynamic Attribute Reduction Based On The Ordered Attributes For Incomplete Information System

Rough set theory is a new mathematical method dealing with fuzzy and uncertainty problem, because it does not need any prior knowledge besides informaton system itself and can classify the knowledge,So it is widely used in data missing and knowledge discovery database. For the data missing, uncertainty, incomplete information system, it has the extremely important theoretical and practical significance to extend the classical rough set theory and form a variety of extension model and knowledge acquisition method.Real word is changing, the information system is changing dynamically. The objects enter and exit the system constantly. So the information system’s original attribute reduction has changed and it is perhaps invalid. At this moment, we cannot use the static attribute reducton algorihtm because it cannot satisfy the requirements for application. We need to modify the attribute order dynamically. Therefore, it has very important theoretical valuable significance to study the dynamic attribute reduction based on attribute order in the incomplete system.During our real life, a dataset is usually shared by multiple users, different users have different needs or interests, they all want to excavate knowledge according interseting. The ordered attributes is an important description of user requirment, it can show the importance of attributes for the user. But in the process of knowledge acquisition, we cannot collect data or make data lost in the process of sorting or storing because of some problems such as user’s privacy issues. So our research object is usually an incomplete information system which has fuzzy and inprecise knowledge. Therefore, how to excavate knowledge from a incomplete dataset for user’s requirment is an important question in datamining.Combine two points above, in this paper, we reasearh the dynamic attribute reduction based on the ordered attributes of the incomplete information system. In the incomplete information system, we use limited tolerance relation to extend the rough set model, and present the new information entropy function and the significance of attribute, form the ordered attributes so that to reflect the use’s need and interest. In order to deal with the problem that the information system has a huge scale, we fuse the idea of decomposition and dispose the pleniful datas. We propose a kind of attribute reduction algorithm based on the ordered attributes.To deal with the problem of update dynamically of the information system in real life, we research the incremental attribute reduction based on the ordered attributes of the incomplete information system, and a reduction algorithm is given. Because most of the existing dynamic reduction is focused on incremental attribute reduction, however, reductive attribute reduction of the incomplete information system is still very immature. So, in this paper we further research the reductive attribute reduction, we analyze four kinds of cases, and propose a kind of reductive attribute reduction algorithm based on the ordered attributes of the incomplete information.