The Research Of Classification Algorithm Based On Rough Set And Neural Network

Recently, with the globalization of information and rapid development of technology, each industry field has accumulated a large-huge amount of data, however, it is difficult to extract useful and important information from a large-huge amount of data. In this case, Data mining technology as a tool of analysis of the data was produced. Data classification is the basis and core of the data mining technology, for it can be used to extract important data description model, predict the future trend of the data and implicit rule, summarizes the data and provide decision strategy for decision makers. Therefore, the stable, rapid and accurate method of data mining classification is not only of great theoretical value but of application value.Classification algorithms commonly include:support vector machine algorithm, the decision tree method and artificial neural network method, etc. As the algorithm is different, the generated classification model is different, the efficiency of data classification and the classification accuracy will also varies because of the classification model. Therefore, choosing the appropriate classification algorithm to process the data quantity of the classification problem is of great importance.In this paper, we combine the rough set theory method with the neural network method for data classification, give full play to taking the advantages of the two complementary to deal with the data of high dimension and large-huge amount of data, etc. Rough set theory can deal with uncertain and incomplete information, without any prior knowledge, through the data itself we can obtain the correlation between the data, and the attribute reduction, but due to the rough set is sensitive to noise, the noise of the environment problems in classification is not accurate; through a certain network topology, Artificial neural network by simulating the human biological neural networks, there will be multiple processing functions of nodes (neurons), in dealing with imprecise, incorrect data and complex nonlinear mapping problem and so on, artificial neural network method has good fault tolerance, adaptive ability and the able to resist noise. But if the high dimension data is easy to cause neural network slow convergence speed, low classification accuracy, the problem such as learning time is too long. Thus this article combine the rough set theory with the neural network, puts forward a new classification model, namely, rough set-neural network classification model, the attribute reduction of rough set as the front-end processor of neural network, firstly using the improved particle swarm optimization algorithm to decision table attributes discretization, then carrying on the attribute reduction based on attribute importance degree, reduce the dimension of attribute space, thus shortening of the neural network learning and training time, improving the accuracy of classification.Finally, the simulation experiments on the actual data is classified prediction effect, which proves that the method is feasible and effective.