Discussion of Some Problems About Nonlinear Time SeriesPrediction Using ν-Support Vector Machine

Abstract

Some problems in usingν-support vector machine (ν-SVM) forthe prediction of nonlinear time series are discussed. The problemsinclude selection of various net parameters, which affect theperformance of prediction, mixture of kernels, and decompositioncooperation linear programming ν-SVM regression, which result inimprovements of the algorithm. Computer simulations in theprediction of nonlinear time series produced by Mackey–Glassequation and Lorenz equation provide some improved results.