Research On Fuzzy Twin Support Vector Machine Algorithm Based On Structural Information

As an effective classification method,twin support vector machine has been widely used in many fields,such as pattern recognition,text classification,network intrusion detection,disease diagnosis and so on.Its goal is to construct two non-parallel hyperplanes by solving two smaller quadratic programming problems,so that each plane is closer to the samples in one class and far away from the samples in the other class.At present,people have carried out indepth research on the twin support vector machine,and proposed different improved algorithms,such as Fuzzy Twin Support Vector Machine(FTSVM),Twin Bounded Support Vector Machine(TBSVM),Least Squared Twin Support Vector Machine(LSTSVM),etc.In order to further improve the performance of the twin support vector machine,this thesis studies the fuzzy twin support vector machine based on structural information and the structural fuzzy support vector machine with parameter margin.In addition,for the proposed model,the model based on least squares is also studied.The specific research contents are as follows:1.The structural information of the sample distribution and the fuzzy information of each sample are introduced into the twin support vector machine,and the structural fuzzy twin support vector machine is proposed.In the twin support vector machine,in order to obtain two non-parallel hyperplanes,the method only considers the interclass separability of the samples,and ignores the structural information of the distribution within the sample class.In addition,the method regards the effect of each sample on the classification surface as equal,but does not consider the different effect of different samples on the decision separating plane,which causes the method more sensitive to noise or abnormal data.In this case,based on twin support vector machine,the structural information of data samples and the effects of different samples are introduced into the twin support vector machine,and a structural fuzzy twin support vector machine model is obtained.In order to further reduce the training time,a structural fuzzy twin support vector machine model suitable for least squares method is obtained by modifying the slack variable term in the model,which is called the least squares structural fuzzy twin support vector machine model.2.The structural information,the fuzzy information of samples and the parameter margin are introduced into the v-twin support vector machine(v-TWSVM),and the structural fuzzy twin parameter margin support vector machine is proposed.Twin support vector machine and its improved algorithms mainly consider the distribution of noise in data sets is uniform,however,in practical problems,the noise depends on the location of input samples,which makes the assumption of uniform noise no longer valid.In addition,v-TWSVM model also ignores the structural information in the sample class and the effect of different samples.For this reason,in view of this kind of problems,structural information,different effects of samples and parameter margin are introduced into v-TWSVM,and a structural fuzzy twin parameter margin support vector machine model is proposed.In addition,by modifying the proposed structural fuzzy twin parameter margin support vector machine,the least squares structural fuzzy twin parameter margin support vector machine model is proposed.3.Aiming at the proposed structural fuzzy twin support vector machine model and structural fuzzy twin parameter margin support vector machine model,least squares structural fuzzy twin support vector machine model and least squares structural fuzzy twin parameter margin support vector machine model,the proposed models are solved by quadratic programming method and least squares method respectively.On this basis,a structural fuzzy twin support vector machine algorithm and the structural fuzzy twin parameter margin support vector machine algorithm,the least squares structural fuzzy twin support vector machine algorithm and the least squares structural fuzzy twin parameter margin support vector machine algorithm are also proposed.At the same time,for the nonlinear case,the kernel function is introduced into the proposed model and solved,and the structural fuzzy twin support vector machine algorithm for solving more complex problems is obtained.4.The performance of structural fuzzy twin support vector machine algorithm and structural fuzzy twin parameter margin support vector machine algorithm,least squares structural fuzzy twin support vector machine algorithm and least squares structural fuzzy twin parameter margin support vector machine algorithm are studied experimentally.By selecting the standard datasets in the UCI and Statlog repository and artificial datasets,the performance of the proposed algorithms are verified by the ten-fold cross-validation method,and compared with the typical twin support vector machine algorithm.