Attribute Reduction And Computing Core Based On Rough Set Theory

With the development of computer technology, and the spread of computers.t h-ere has been a explosive growth in date. Whether it is the breadth of data, or d ata accuracy,compared with the past, have a qualitative leap. People can use the past data to guide future production and life.For example, companies can use past data to guide, product manufacturing,transportation, sales, in order to avoid the b usiness risk; state data can be used topredict natural disasters, such as earth qua kes, typhoons,to avoid people’s life and property is threatened; medical treatmen t according to the patient medical data, disease diagnosis, risk assessment of pers onal illness.But on the other hand, due to the increase in breadth and accuracy o f the data for data processing Make it difficult, using conventional means it is dif ficult to timely treatment of these huge data, especially when there are some timeli ness requirements for data processing, such as earthquake prediction. At this tim e we need to adopt a technology,remove redundant data in the database, retrievin g the database can represent thecharacteristics of the data, as the representative o f the database, in order to improve theprocessing speed, meet the real-time requi rements.The main work of this paper:This paper introduces the concept and attribut e reduction research background and research situation of rough set, and the com mon computing core, attribute reduction algorithms are introduced and compare d, highlighting the advantages and disadvantages of various algorithms.Based on the improved discernibility matrix and calculation method,this algori t-hm firstly by modifying the decision attribute values, generating consistent deci sion tables, andthrough theoretical derivation, proved that the decision consistenc y table generation and original table is equivalent in the nucleus, and then propose a decision tabledecomposition algorithm, the original decision the table is divided into two sub table, on the table to add the value of the property, to ensure the s ub table of the same classification ability, and through theoretical derivation of t wo sub sets and table of the original table core set is equivalent to the two sub ta ble, respectively for nuclear, and finally through the UCI data set of experiments prove the validity and accuracy of thealgorithm.In the common attribute reduction algorithm, especially based on the heuristic attribute reduction algorithm for attribute reduction, relative granularity, unable to deal with theproblem of inconsistent decision table. A heuristic attribute reductio n algorithm based on relative granularity is improved, the algorithm is not only suitable for consistent decisiontable, the inconsistent decision table is also applica ble. Finally, the experiments with UCI data sets prove the correctness and validit y of the algorithm.