Study And Application Of Grey Support Vector Machines

For small sample of data information processing, At present, many scholars do considerable researches about gay Support Vector Machine models and made out some achievements, gay Support Vector Machine models can make up the lack of only using anyone of them, and have better data processing and prediction, So it has become a very important topic.At first, the thesis states the characters of small sample data, and analyzes the complexity and the particularity of estimating small samples, then combines gray system theory and support vector machine to establish models.Based on the error analysis and machine of the two existing GM(1,)model, the thesis optimized parameters of background value by particle swarm optimization algorithm,when original series strictly follow the exponent law, then combining the advantages of SVM, a new forecasting model of grey support vector machine based on PSO was proposed. The results of experiment show that the forecasting model is propitious to approximate non-homogenous exponential sequence modeling, and offers a new way to improve the forecasting accuracy, when original series doesn't strictly follow the exponent law.Owing to the complexity of industrial processes, such as multivariable, nonlinear, time-varying, a large number of studies focus on the historical data of industrial processes, using a variety of nonlinear function method to establish models, However, autocorrelation and cross correlation of the input variables would decreasing the generalization ability of the model, therefore, the thesis would take gray relational analysis method as a property of pre-processor, and modify the weight of each factor according to gray correlation, finally, the support vector regression machine was trained to get the optimal structure. The results of experiment show that the model have better generalization ability, is effective and reliable.