Adaptively predicting time series with local v-support vector regression machine

In this paper, we introduce v-support vector regression machine into the frame of local modes in order to obtain the accurate prediction of time series. And to circumvent the vexing problem choosing the number of neighbors by the leave-one-out cross validation error in the local model, we propose an adaptive method based on the estimation of generalization error by using theoretical bounds. The experiments on sunspot data set demonstrate that local v-support vector regression machine gives promising result and the adaptive method to choose the number of neighbors is effective and efficient.