Extension Of Twin Support Vector Machine In Clustering And Fuzzy Semi-supervised Classification

With the advent of artificial intelligence in recent years,machine learning as the core of artificial intelligence and its implementation has also been fully developed.Machine learning uses various algorithms to train a large amount of data to learn how to accomplish a task and make the machine more intelligent.Clustering analysis and semi-supervised classification analysis,as one of the important learning tasks in machine learning,have received a lot of attention from researchers.In this paper,we propose a new clustering and semisupervised classification algorithm to solve the current problems in plane cluster analysis and semi-supervised classification analysis,respectively.The specific research work is as follows:In terms of cluster analysis,plane-based clustering methods measure withincluster or between-cluster scatter by linear,quadratic functions or some other unbounded functions,which are sensitive to samples far from the cluster center.Therefore,we introduce bounded ramp function into plane-based clustering and propose a ramp-based twin support vector clustering(Ramp TWSVC).Since the within-cluster or between-cluster scattering of Ramp TWSVC is measured by the bounded ramp function,Ramp TWSVC is robust to samples far from the cluster center and it is easier to find the intrinsic clusters.The nonconvex programming problem in Ramp TWSVC is efficiently solved by the alternating iteration algorithm,and its local solution can be obtained theoretically in a finite number of iterations.In addition,we propose and solve its nonlinear manifold clustering form by kernel trick.In the aspect of semi-supervised classification,aiming at the defects of existing semi-supervised classification algorithms,that is,how to use unlabeled samples safely to prevent the performance degradation of classifiers,we extends the point-based intuitionistic fuzzy number to a plane-based intuitionistic fuzzy number and introduces it into a twin support vector machine for semisupervised learning,and proposes a semi-supervised intuitionistic fuzzy twin support vector machine.In SIFTSVM,the plane-based IFN is organically incorporated into the twin support vector machine,which safely learns unlabeled samples through an asymptotic process and is highly adaptive to unlabeled samples and noise.A series of mixed integer programming problem is considered in SIFTSVM,and we propose a convergent alternating iteration algorithm to obtain an admissible solution that achieves stepwise learning of unlabeled samples and dynamic updating of the classifier.Furthermore,SIFTSVM is extended to the nonlinear case by a kernel trick.In this paper,simulation experiments are conducted on a large amount of artificial and benchmark data,and the experimental results on several artificial and benchmark data sets show that the proposed Ramp TWSVC and SIFTSVM have better performance and are more competitive than the related plane-based clustering methods and semi-supervised classification methods.