Research On Data Reduction Based On Extending Of Rough Set

Rough set theory is a new mathematical tool dealing with vagueness and uncertainty, has found its applications in many areas such as AI, KDD, pattern recognition and classification and fault diagnostication. Attribute reduction is an important part of the RS theory, which deletes the redundant attributes on the premise of unchanged of classification precision. And how to find the reduction from the huge and complexity data set quickly and efficiently is a key step of RS theory.Attribute reduction is focused on in this dissertation. Main topics include1. A new algorithm FFGAR is presented based on branch and bound to compute U/A.Using the including relationship,the algorithm cuts down the complexity to O (|A||U|). And making use of the property that positive region increases with the amount of attributes, a fast algorithm whose computation complexity is O(|A|~2|U|) is proposed.2. In most of the exist algorithms, the dependency function is employed to evaluate the quality of a feature subset. The disadvantages of using dependency function are discussed in this paper, and the problem of the forward greedy search algorithm based on dependency function is presented. A new consistency measure is introduced to deal with this problem. Based on the consistency, a forward greedy search algorithm GARBC is designed to find reducts.3. Classical RS model just work in discrete spaces, so we need transform numerical attribute to discrete ones. The process may loss some information of the data set. To deal with this problem, neighborhood rough set model based on neighborhood system is introduced and a new rough set model with the mixed attributes data set is constructed. Based on the mixed consistency, a forward greedy search algorithm ARBMC is designed.4. Experiment results with UCI data sets show that the proposed algorithms are efficient and effective. In dam's safety monitoring system, we apply the method based on RS theory to analyze the data of dam. The practice results show that it can simplify the evaluating of the new data, and finding useful information from existing data. The results express that the method can remain the classification power and reduce the computation time.