Applied Research Of Support Vector Regression Machine Based On Parameter Optimization

The manufacturing approach of micron wood fiber mould is a highly complicated process. This process has the characteristic of such as nonlinear,time-delay,high-dimensional and etc. Mould pressing products has high density and it can be used to further processing and decorate, so it has a high application value.Holding nail force is an important indicator for measure the physical properties of materials.When the mould pressing products are connected by screws, the magnitude of holding nail force is more important.Therefore,predicting the holding nail force of mould pressing products becomes an important subject of manufacturing approach of micron wood fiber mould.This paper view the shift-handle of automobile as a target,and bring the machine learning into the forecast of the holding nail force of mould pressing products. Through the machine learning method it finds an effective way to predict the holding nail force.Firstly,this paper introduces the process of mould pressing.This can provide basis for the eigenvector selected of the prediction model and the establishment of the prediction model. Then,it studies the BP network which is very popular in prediction to establish a prediction model to predict the holding nail force.Through experiments,the results indicate that although this approach has higher forecast accuracy,there are some problems such as the structure of model is difficult to confirm,over learning and when using gradient descent algorithm,the model is easy to be trapped into local minimum and so on.Statistical Learning Theory is based on a solid theoretical foundation.It provides an unified framework for solving the small sample learning problem.Support Vector Machine(SVM) is a machine learning method based on Statistical Learning Theory.It can solve a series of issues of Neural Networks.Compared with BP network,SVM has better robustness and higher forecast accuracy.It is an effective machine learning method.Finally,the paper establishes a prediction model using Least Square Support Vector Machine(LS-SVM).And to solve the problem about the difficulty of parameters selection of SVM,the paper studies the method which use Particle Swarm Optimization Arithmetie(PSO) to optimize the parameters of SVM.This method can implement the parameters of SVM to be selected automatically.This paper proposes a LS-SVM regression model optimized by PSO and uses this model to forecast of holding nail force.The experiment results indicate that the PSO can autoly search the parameters of SVM.The LS-SVM prediction model which optimized by PSO has higher forecast aecuracy than BP network prediction model,and more suited to the forecasting of holding nail force.