Soft Rbf Neural Network Models and an Svm Model for Wages Time Series Modeling and Forecasting

The wages time series are in fact stochastic in which successive observations are dependent and can be represented by a linear combination of independent random variables , , 1 − t t ε ε ... . If the successive observations are highly dependent, we should use in model past values of the time series variable and (or) current and past values of the error terms { t ε }. There are available techniques which are designed to exploit this dependency and which will generally produce superior forecasts. Many of these techniques are based on developments in time series analysis recently presented by Box and Jenkins. This article offers computational algorithm used in SVM method and in soft RBF neural networks and conducts experiments using these algorithms for wages time series modelling. Section 1 firstly describes the framework of SVM ́s methods and support vector (SV) regressions and briefly describes construction of NNW for the comparison and verification SV regression wages forecast. Section 2 introduces the soft RBF neural network. In Section 3 the approximation and prediction abilities of SVM method is compared with the soft RBF NNW approach. A section of conclusions will close the paper.