Support Vector Machine For Solving Classification Problem And Its Improvement Strategies

Support vector machine is an important method in machine learning and a tool used to solve classification problems.In 1995,Vapnik proposed support vector machine based on the principle of structural risk minimization in statistical learning theory,and since then,support vector machine has attracted extensive attention.Then,the model of constructing hypersphere support vector machines is proposed,which broadens the research direction of support vector machine from single hyperplane to hypersphere.However,in the face of large sample data,the classification efficiency of support vector machine is affected to a certain extent.Based on the existing research results of support vector machine,this paper studies the hypersphere support vector machine and the acceleration algorithm of support vector machine.The main contents are as follows:Chapter 1 is the introduction,which introduces the background of machine learning and the history of support vector machine,discusses the advantages of support vector machine compared with traditional algorithms,and briefly describes the idea of standard support vector machine model establishment and solution method.Finally,some preliminary knowledge is given.Chapter 2 presents a kind of preprocessing algorithm for data samples.Based on the feature of less support vectors in support vector machines and the clustering property of samples,three preprocessing algorithms are proposed,and the effectiveness of the algorithm is proved theoretically.Then a numerical experiment was carried out,and the experimental results showed that under a certain reduction ratio,the preprocessing algorithm could maintain the classification accuracy of SVM and reduce the solving time of the model.In Chapter 3,a modified hypersphere support vector machine is proposed.A new modified hypersphere support vector machine model is proposed based on the preprocessing algorithm and the relationship between data samples.The detailed solving process of the model for dichotomy problem and the corresponding discriminant method are given,and the relationship between the model’s support vector and the hypersphere radius is proved theoretically.Finally,numerical experiments show that the proposed model has better classification performance than the standard hypersphere SVM model.In Chapter 4,the modified hypersphere support vector machine model proposed in Chapter 3 is extended to the multi-classification problem,and the prior knowledge improved algorithm is introduced to convert the quadratic programming problem into the primary programming problem,which solves the defect of the high computational cost of the model.Finally,numerical experiments are carried out to verify the effectiveness of the proposed algorithm in dealing with multiple classification problems.Chapter 5 summarizes the research work of this paper,and puts forward some ideas and problems to be solved in the future.