Research On Attribute Reduction Algorithms For Data Mining Based On Rough Set

Data Mining is a new hot research spot in database area. It is used to discover the implicit, previously unknown, and potentially useful information from the large quantity of data. Rough Set methodology is a kind of more valid method to deal with it. Attributes reduction is the most important step in the course of data mining based on Rough Set theory. How to acquire a simple and effective attributes reduction method has a great meaning for Data Mining domain.Currently, there are lots of algorithms for attributes reduction, in which the algorithm based on discernibility matrix is one of efficiently algorithms. Aiming at the drawbacks exist in the traditional attributes reduction algorithm based on Skowron discernibility matrix, a novel algorithm based on enriching boolean matrix is presented which include two innovational points as below: Firstly, the concept of enriching boolean matrix is proposed, which can effectively cut down the computing and storing capacity. Secondly, the generation algorithm of minimal disjunctive normal form (DNF) for discernibility function is presented, which can directly generate the minimal DNF of discemibility functions and saves memory space and CPU occupation time. Finally, the validity and effiectiveness of the attributes reduction algorithm based enriching boolean matrix is proved.Aiming at the situation in real application which data are always changing in database, an incremental algorithm of attributes reduction is proposed. This algorithm can avoid reduction from the large original decision table when new objects are added. It can update and vindicate the results of reduction for original table, and improve the efficiency of attributes reduction for database. The algorithm is demonstrated to be effective with the specific example ultimately.