Galkin O. Extensible hyperspheres technique on the basis of support vector machines for classification problems

It has been proposed and investigated an extensible hyperspheres technique for solving classification problems which is based on support vector machine method and its main property of the continuous decrease of the empirical classification error and increasing of the margin. There is developed a new approach to the determination of decision rules of the classification problems based on the construction of an optimal hyperplane which is presented in the form of the curve increasing efficiency and reducing errors in the solution of practical classification problems. In this work it has been defined a property of extensible hyperspheres technique based on support vector machine to operate effectively with any number of classes, as well as to determine the probability of belonging a new point to the class.