Research On Non-parallel Support Vector Machines Based On Optimization Problems

Support vector machine(SVM)is one of the most widely used algorithms in machine learning.It was first put forward by Vapnik in the 1990 s,along with the rapid develop-ment of machine learning and data mining.Support vector machine(SVM)classification and regression problems new meaning,attracted the attention of more and more scholars and research.It is widely used in image processing,speech recognition and natural lan-guage processing,etc.The traditional support vector machine(SVM)is by the principle of maximum interval for single decision hyperplane to achieve the purpose of classification.But its generalization is limited by a single hyperplane.The classification of cross date and other date is not strong.As twin support vector machine(SVM)and the parallel support vector machine(SVM)are proposed,the corresponding nonparallel decision hyperplane is constructed respectively for each type of sample points.Their generalization and flexibility are stronger.We are in a big data era background.As the data types of diversity and particularity,the size of the data and the requirement of data classification models are im-proving.In this paper,from the angle of the optimization problems and the classification of the multiple points of view,the author sets up new non parallel support vector machines(SVM)classification models and shows their effectiveness by numerical experiments.The structure of this paper is organized as follows:In Chapter 1,the author introduces the background,significance,research status and the main job of the paper.In Chapter 2,based on the nonparallel support vector machines(NPSVM)algorithm,a class of the new classification algorithms is proposed.The author discusses the optimization problem of l1-v-not parallel support vector machines(l1-v-NSVMOOP).A new optimization classification model makes the algorithm have good classification performance to solve the problem of unbalanced classification.Numerical experiment results show that the algorithm is effective for dealing with unbalanced classification problem.In Chapter 3,based on the twin boundary support vector machine(TBSVM),a class of the new classification algorithm for multiple points of view is proposed.The author studies the multi-view twin boundary support vector machines(MvTBSVMs).A new model of opti-mization classification is established.Connections and differences between multi perspective date are fully utilized.It has good classification performance and advantage in dealing with date on multiple points of view.The author proposes the corresponding algorithm.The numerical experiment results show that the effectiveness of the algorithm.In Chapter 4,the research contents of this paper are summarized,and the directions for further research are proposed.