Vessel Purchase Cost Time Series Forecasting Based on C-SVM
Kaigui Xie · Journal of Academy of Equipment · 2012
Vessel purchase cost forecasting faces two problems: small samples and many outliers.Although support vector machines(SVM) is able to forecast small samples,it is too sensitive to fit outliers;fuzzy support vector regression improve that,but the outliers classifactory lack of fuzziness.Using cloud theory which can express fuzziness scientifically,through the improving of backward cloud generator,the cloud membership generator for fuzzy classify of outliers is designed;by transferring the cloud membership into SVM,the algorithm of cloud membership-based SVM(C-SVM) is presented.On this foundation,the model of vessel purchase cost forecasting based on C-SVM is founded.Simulation results prove that C-SVM can not only depress the sensitive of model for outliers fuzzily,but also find the better bound level of support vectors adaptively.The precision of vessel purchase cost forecasting is improved by this model.