Research On Improvement And Application Of Support Vector Machine Algorithm

Support vector machine is a machine learning method based on statistical learning theory.Statistical learning theory includes VC dimension theory and structural risk minimization.Support vector machine training algorithm is essentially a quadratic programming problem and quadratic programming theoretically has a global optimal solution.Therefore,it is widely used in network traffic classification,data mining,signal processing and many other fields.This paper mainly studies the theory,kernel function,support vector machine parameter selection method of support vector machine algorithm and application of support vector machine in network traffic data classification.The specific work of this article includes the following three aspects:(1)In the traditional support vector machine,the distance between the optimal classification hyperplane and the two types of samples is equal,which leads to the classification hyperplane tending to the type with dense data set distribution or large sample size.For this problem,this paper proposes a structured support vector machine based on unequal distance classification hyperplane margin.This method introduces the idea of??unequal interval into the support vector machine,which makes the classification hyperplane have better classification performance.In addition,we introduce the local structure information between samples in this method to enhance the support vector machine Learning performance.Finally,it is verified on the UCI dataset that the method can effectively improve the classification accuracy of the model.(2)This paper studies the construction method of the Fourier kernel function,and we have obtained a new method of replacing the cosine term in the Fourier kernel function by using quadratic polynomial function.At the same time,we verified that the new function is an effective kernel function.Experiments show that the new kernel function can achieve the same accuracy as the Fourier kernel function and the calculation efficiency is faster.(3)The parameter selection in the support vector machine often affects the performance of the classifier.Today,intelligent optimization algorithms are a more commonly used method of parameter selection.Therefore,this paper proposes a new hybrid algorithm to select the parameters of support vector machines,which effectively improves the generalization ability of support vector machines.And three kinds of support vector machines are applied to the classification of network traffic data,and we compare and analyze the performance of three kinds of support vector machines.