Research On Data Mining Methods Based On Attribute Reduction Of Rough Set And Optimization Theory

Data mining is an important method of data analysis of system and informationprocessing, of which the core issue of the research is to establish the mining model. For thesake of comprehensive analysis of the problem, it often requires to use a lot of system-relatedvariables or factors. On account of numerous attribute variables, it results in difficulties inmathematical modeling of system and properties analysis. In fact, the simulation model of thereal system is mainly used for simulating major changes of the system and reflecting theimportant characteristics. Therefore, in the circumstance of lacking of prior knowledge, it hasimportant implications of using rough set attribute reduction method to determine the coreattributes of the system, as well as filtering out unwanted attributes and creating a miningmodel and combining with the evolutionary optimization technique for practical application.This paper elaborated and analyzed the basic method and the mining model andapplication technology of data mining firstly, as well as basic theory of rough set attributereduction method and intelligent data mining modeling techniques. On the basis of the studyof classical rough set attribute reduction algorithm, we combined with the characteristics ofevolutionary algorithm, and respectively established two kinds of rough sets attributereduction methods which were separately based on genetic algorithm and particle swarmalgorithm, and we also analyzed the nature of these two methods. On the issue of patternmining and diagnostic analysis of the problem, we respectively used artificial neural networksand procedure neural networks to establish mining models, and proposed a training algorithmof procedure neural networks which was based on optimal piecewise approximation. On theissue of reservoir evaluation which was based on logging data, we preprocessed the dataattributes and data using the attribute reduction optimization algorithm which was establishedby this paper. This helped to reduce the data dimension which was used for modeling andreduce the information redundancy to provide a data source of a relatively high quality forcreating the mining model. It obtained a relatively good application result by constructing adata mining model which was based on the rough set attribute reduction and procedure neuralnetworks.This paper proposed a method of attribute reduction which was based on thecombination of rough set and optimization theory. And it could effectively simplify theredundant attributes and reduce the data dimension of the model and computationalcomplexity. It improved the function approximation capabilities which were based on the artificial neural networks and the procedure neural networks, as well as the modelingflexibility and adaptability. Using it as the mining model and combining with the attributereduction pretreatment method, had achieved the information mining of reservoir lithology,reservoir fluid properties and reservoir parameters of property which were based on a largeamount of data in logging information. It provides a new method for the research of reservoirevaluation basing on logging data of oil field, and possesses a relatively big application value.