Research On Improved Non-parallel Support Vector Machine Algorith

Support vector machine(SVM)is a new supervised pattern recognition method in the field of machine learning.It has good ability to process and predict unknown samples.SVM was introduced by Vapnik in the 1990 s.Due to its solid statistical learning and optimization theory foundation,it has received much attention from scholars at home and abroad.SVM plays a great role in dealing with small samples,high dimension and nonlinear problems.It has the characteristics of high fitting accuracy,strong generalization ability and global best.The traditional machines learning method adopts the principle of empirical risk minimization to achieve the purpose of training,while the support vector machine is the principle of structural risk minimization on the basis of ensuring the maximum interval.It uses the limited sample data information to find the optimal balance between the complexity of the model and the learning ability,so that the model has a better generalization ability.In traditional support vector machines,two parallel support hyperplanes were constructed by a quadratic programming problem,which guaranteed the maximum spacing between hyperplanes.The intermediate hyperplane was taken as the final decision hyperplane.The twin support vector machines,nonparallel support vector machines and a series of support vector machine models derived from them all construct a pair of nonparallel hyperplanes by solving two small-scale quadratic programming problems,so they can realize effective classification of data samples.With the continuous update and development of database,the classification accuracy of data has higher requirements.The existing model should be improved accordingly to achieve better classification effect and promotion ability.In order to improve the classification accuracy and operation efficiency,the author improves several existing models and establishes three kinds of nonparallel support vector machine models,namely -improved nonparallel support vector machine,improved sparse soft interval nonparallel support vector machine and -soft interval non-parallel support vector machine.The corresponding classification algorithms are given.Numerical experiments are conducted to explain the effectiveness of the new models.The detailed arrangement of this paper is as follows:In Chapter 1,the author mainly introduces the relevant background,important models,research progress of support vector machine and a brief introduction to the work carried out in this paper.In Chapter 2,an improved ν-nonparallel support vector machine is proposed.Based on nonparallel support vector machine,the parameter constraint of is applied in the objective function.On the basis of achieving the classification optimization,the small band is maintained.At the same time,the regular term is added to minimize structural risk.The parameter which has no real value meaning is replaced by .The support vector can be effectively controlled by optimizing the value of .The effectiveness of the new model is explained by numerical experiments.In Chapter 3,an improved sparse soft interval nonparallel support vector machine is proposed.Based on the soft interval ε-band nonparallel support vector machine,soft interval secondary loss function and ε-band insensitive secondary loss function are introduced to reduce the complexity of the model.While ensuring the sparsity,it can effectively predict the unbalanced data problem and have good classification ability.In Chapter 4,a ν-soft interval nonparallel support vector machine is proposed.By combining the soft interval ε-band nonparallel support vector machine model with the ν-support vector machine model,a corresponding classification algorithm is proposed.The algorithm improves the classification speed and has a better classification effect.The effectiveness of the proposed algorithm is explained by numerical experiments.In Chapter 5,the research content of this paper is summarized,and the future research direction is further analyzed.