| Support vector machine (support vector machine, SVM) is proposed a new algorithm based on statistical learning theory.SVM find the unknown probability distribution function by a special mapping. Moreover, it has been established as a standard tool in machine learning and data mining. SVM approach avoids that the multi-layer feedforward neural network structure is difficult to determine, over-learning , less learning, local minimum and so on. Currently,SVM is considered to be the best theory which learning machine is promoting performance issues for the "small sample set," or "small sample size".At present,Support Vector Machine has been widely used in a lot of practical problems. However, the performance of SVM is closely related to the selection of parameters. For existing parameter selection method, Calculated quantities is a very great number and promotion is very bad. In practice, the operation is complex, and the results are not good at all. On the other hand, research direction is the kernel function itself. Support Vector Machine is fully described by a core function and training set. Kernel function is the core of support vector machines, different kernel functions will result in different classification results. Common kernel functions have their own advantages and disadvantages. The key of improving SVM performance is how to build strong learning ability and generalization ability of the kernel function. In this paper, using intelligent algorithms to optimize the SVM parameters acquire to be more accurate parameter values in practical applications for classification and fitting. So, this method can overcome the randomness and blindness of SVM parameter selection,which greatly improved the performance of the SVM. At the same time, in order to construct the kernel function for a given problem,the paper designs a new nuclear function - mixed kernel function, with good learning ability and generalization ability through a large number of experiments and proved. |
PDF Full Text Request