Wave-matching based SNURBS for time series prediction

Time series prediction is widely applied in the field of signal processing, economic, weather and so on. Making the most of time series to predict the future is a hot issue. Traditional ways only use time series with a certain error, and forecast the future states at some moments. Motivated by making full use of time series, the NURBS expression with time parameter (SNURBS) is utilized to model time series. SNURBS can describe the explicit function between the system behavior and the time. Furthermore, the algorithm of wave matching is used to predict control points of the future behavior curve, and by adding them into the SNURBS we can realize the prediction of the continuous behavior in a period of future time. This new method is called Wave-Matching Based SNURBS (WM-SNURBS) in this paper. To prove the forecasting ability of our method, we exploited simulation experiments about Lorenz system under several conditions, and experimental results show that WM-SNURBS can effectively predict time series.