Application Of Neural Network Rough Sets In Data Mining Based On

In the early1980s, the United States futurists John Naisbitt in his first book"Megatrends" mentioned that human beings were flooded by information, but were thirst forknowledge. The rapid development of computer hardware provides a large number of datacollection tools and storage equipment for human, the mature of database technology hasmade the data that human accumulated tends to linear growth with "J" type curve. Theemergence and the development of the Internet technology has made the world as a whole,people can exchange information and work on the Internet through time and space. In thisinformation-overloaded age, facing the vast information, people need to express theknowledge that changed from huge data by discarding the dross and selecting the essential.Data Mining (DM) arises in such background.The common algorithm and theory of Data mining contains Rough Sets theory,Artificial neural network, Decision tree, Genetic algorithm, Classification, Clustering, etc.Rough set theory was put forward in1982by Z.P awlak, the tool first determines theapproximate interval which not given certain features or property situation through theindiscernibility relation, then determines the relationship between internal attributes. Indealing with large data and eliminating the redundant information, etc., rough set theory hasa very good effect. So, rough set in the data mining application has a wide developmentprospect and application value. However, because the rough set theory is too simple todescribe the certainty of error, it is sensitive to the noise of the object.As artificial neural network has such advantages as strong robustness, highclassification accuracy, insensitive to the noise data and so on, it has been widely used inmachine learning, pattern recognition and other fields. However, facing the problems ofhigh dimensional and very large scale in data mining, the neural network’s defects such asits very slow learning speed, easy causing network overtraining, poor rules generation aremore apparent, the original neural network algorithm will face some problems in efficiencyand extensibility.Faced with these problems, this paper puts forward a new method of data mining,combines RBF neural network with rough set theory, applying to large database miningclassification rules. The BP neural network in artificial neural network has been widelyapplied in data mining, but in practice, compared with neural network data preprocessing method, we found that RBF neural network has faster convergence speed, higher accuracyand more reliable. Therefore, this paper will use RBF neural network for the data training.The main idea is using the advantages of RBF neural network through the network trainingand learning to eliminate repetition data and interference data, transfer the processed data torough set for further attribute reduction and rule extraction, then get the final miningknowledge.This paper will apply this data mining method which combined RBF neural networkwith rough set to parameter data mining of high and new technology enterprise indevelopment zone, and at the same time. It compares the data mining result of rough setmethod with that of not using neural network to prove its superiority and validity.