Several Rough Models And The Combanation Of Rough Set And Neural Network Research

Rough set theory provides mathematical methods, and it also has applications in real life.Thistheory deal with the main problems including data reduction in database, data correlation, similarityand differences in the findings. This theory makes reduction summarize to get commonness andcharacteristics, and this method can be applied to the classical rough set some promotional model. Itprovides an important theoretical value for the in-depth study of the classical rough set. And itextends the classical rough set applications range.But rough set is limited, such as sensitivity to noise data, so the combination of rough sets andother intelligent methods research has become a hot research spot. Neural networks are extensivelyinterconnected to form a complex network system by neurons. This article describes the combinationof rough set and neural network research. Due to the characteristics of rough sets and neuralnetworks, there are a lot of methods to achieve a comprehensive way. From domestic and foreignresearch results, the combination of rough set and neural network has several typical methods:(1)Rough sets as a front-end processor of the neural network is the most common way.(2) Roughneural network.(3)Neural network is built with rough set theory. In this paper, the main results aresummarized as follows:1. The relation of rough fuzzy set and fuzzy rough set is studied. Rough fuzzy set is extended togeneralized space, and the relationship between the upper and lower approximation in generalizedrough fuzzy set model is proved. The nature of rough membership function is studied in ageneralized approximation space.2. To do two important expansions for fuzzy rough set: At first, fuzzy rough set is expanded tothe intuitionistic fuzzy environment. The concept of intuitionistic fuzzy cut set is proposed inintuitionistic fuzzy rough set. The concept of upper and lower approximations in intuitionistic fuzzycut set rough set is proposed and some important properties are proved. Then an intuitionistic fuzzysimilarity relation is proposed and the classic attribute reduction algorithm is applied in this similarrelationship. An example is proposed to demonstrate that common attribute reduction algorithm isapplicable to the new proposed rough set and extends the range of applications of classical rough set.Secondly variable precision rough set is expanded to the fuzzy environment and form a variableprecision fuzzy rough set. The relationship between this variable precision fuzzy rough set with the first part of fuzzy rough set is studied and proved. The attribute reduction algorithm based on theaccuracy of the approximate is proposed. An example is proposed to prove the correctness of thisalgorithm.3. The research and the combination way between neural network and rough set are summarized.Then a simulation example is proposed to illustrate that the combination of the two theories ismeaningful because it can simplify the structure of the neural network, shorten the training time.