Multi-parameter Adaptive Prediction of Chaotic Time series based on LS-SVR

Nowadays, fault detect and prediction is quite important for the purpose of ensuring the correct functioning of complex system; nevertheless, it is usually difficult to establish an exact mathematical model in analytical form for complex system, therefore, fault prediction of complex system always relays on the analysis of the observed chaotic time series. In order to enhance the validity and accuracy of the prediction process, all relevant multi-parameter chaotic time series information is taken into consideration in this work. Then, multi-parameter phase space reconstruction process is performed to generate training samples; and a multi-parameter adaptive prediction model using least squares support vector regression approach is established in the end. The proposed method is based on the support vector machine prediction theory. In this manuscript, the simulation experiment of chaotic time series with three parameters of certain equipment is investigated and presented for an illustration. As indicated by the results, the proposed method is of good prediction accuracy; furthermore, it is shown to be an effective prediction method.