Research On Support Vector Machine Algorithm And Its Application In Financial Crisis Warning

In this paper,we mainly study the support vector machine(SVM)algorithm.On the basis of the previous use of parameter optimization to improve the accuracy of SVM classification,we study how to reduce the training time of SVM by reducing the training data set.In order to make the training speed and classification prediction of SVM faster and more accurate,we make the following improvements and generalizations to SVM algorithm.1.Support vector preselection algorithm based on distance pairing sortingAccording to the distribution characteristics of support vector and its importance to SVM classification decision function,we propose a support vector preselection algorithm based on distance pairing sorting(DPS).Using this algorithm to preprocess the training data set can construct a support vector candidate set which contains all the support vectors as much as possible,and then let the support vector candidate set represent the training data set for SVM training.The SVM based on DPS support vector preselection algorithm can keep good classification accuracy and shorten a lot of training time.2.Support vector machine algorithm based on DPS support vector preselection and parameter optimizationAlthough SVM algorithm based on parameter optimization has good classification accuracy,its training time is usually very long.In order to solve this problem,we introduce DPS support vector preselection algorithm into the algorithm,and then propose a new algorithm.That is SVM algorithm based on DPS support vector preselection and parameter optimization.Through the complexity analysis and experimental results of numerical simulation,the new algorithm has less computation and shorter training time.3.Using new SVM algorithm to solve the financial crisis early warning model based on SVMThe financial crisis warning problem can be transformed into a binary classification problem based on SVM,which can be solved by SVM algorithm.The SVM algorithm based on parameter optimization proposed by predecessors has better classification accuracy.We introduce DPS support vector preselection algorithm into the algorithm and propose a new SVM algorithm.We choose the financial datas of specially treated(ST)enterprises and non ST enterprises to carry out comparative numerical experiments on various SVM related algorithms.The results show that the new SVM algorithm proposed in this paper maintains the high classification accuracy of SVM algorithm based on parameter optimization,but its training time is relatively reduced by about 50%.The new algorithm improves the application efficiency of SVM in financial crisis warning.