Research And Application Of Data Classification Method Based On Rough Set And Neural Network

Now days, with the rapid development of information technology and database technology, data has been an absolutely necessary part in people's work and life, but it is a very difficult task to extract the useful information from massive data. Data classification is an important data analysis technology, and it can be used to extract the model that describes important data category and predict the future data trend, then discover the hidden rules to supply service for business decision and strategic development.In this paper, the rough set theory is introduced firstly, and an reduction algorithm based on the important degree of attribute is imported, which can remove the redundancy information effectively and reduce the dimension of input. Secondly, a neural network classifier based on dynamic threshold is proposed aim at multi-class problem, and this classification model a different threshold function from traditional ones, which improves the classifier's generalization ability effectively. In addition, the optimization design of the structure parameters such as the number of hidden layers and their own nodes is operated by combining theory guidance and experiment, and then the BP neural network with two hidden layers is employed as the kernel of the classifier model. Finally, the rough set and neural network are combined by making up for each other's deficiencies, and the rough is used as the front processor of the neural network model to reduce the input vector for the purpose of select variable scientifically and improve the performance of the model.Box-office prediction is a typical nonlinear problem, and can also be translated to a classification problem. All the movies are divided into six categories according to their box-office incomes, and the purpose is to predict a film into the right class. On the basic of successful modeling, factors that effect the box-office revenue are selected as the input variables, and their initial values are determined by using statistical method. Then the proposed classifier model is used to solve this problem. According to compare the results with decision tree, support vector machine, and RBF neural network, the neural network classifier with dynamic threshold has better forecasting accuracy, stability and generalization ability. The combine of rough set and neural network can obtain favorable effect to improve the classification performance, and distinctly better than other models.