The Research Of Data Reduction And Application Based On Rough-set

Rough Set theory as a mathematics tool of dealing with fuzzy and uncertainty knowledge,has a wide range of application and reasearch in the area of data reduction, for the issue of thetraditional rough set theory does not support in the reduction of incomplete decision table,extended rough set theory have been proposed, and similarity relation, tolerance relation andlimited tolerance relation etc have development form the equivalence relation of the traditionalrough sets. currently, reduction algorithm based on rough set theory, such as identification matrixreduction algorithm, attribute significance reduction algorithm and so on, are based on thereduction of data but ignore the relationships between data, at the same time, due to theminimum reduction is obtained a NP hard problems theoretical, the reduction efficiencydecreased dramatically in the reduction of large and complex data sets, which leads to thereduction algorithm is applicable and operable.On the foundation of have analyzed the existing reduction algorithm based on rough settheory, in order to respond the questions which faced currently, this paper proposed a clusteringfeedback reduction algorithm models. It’s proposed an attribute reduction algorithm aiming atthe problem of the traditional reduction algorithm ignoring the relationships between data,describe the relationship between data through extraction contact point, combine the randomnessand fuzziness through cloud model theory, divide the data set based on cloud model, combinedwith the thought of granular computing, the algorithm is design into a parallel algorithm, and byestablishing virtual node construct clustering granulation tree, nearest neighbor transmission asthe strategy, feedback fixed as the guidance, makes the algorithm meet the requirements ofincremental reduction. When constructing virtual node, through the adjustment of themembership degree classification rules, Increase reduction efficiency. Introduced the concept ofdata integration mapping at the extraction of data point and associated point to solve problemswhich caused by the platform difference.The clustering feedback reduction algorithm models based on data relationship in this paperbased on the conclusion of previous studies, borrow ideas from the characteristics of varietyalgorithms, at the same time using hierarchical thinking, the reduction process has beenimproved. Simulation experiment shows that this algorithm model is feasible and effective.