Local linear approximation methods are one of the common tools for nonlinear prediction of chaotic time series. In the case of predicting time series data corrupted by observation noise with this method, it is possible that unsuitable points are selected as spurious neighbors, which causes lower prediction accuracy. To solve this issue, we propose a simple noise reduction method, which is based on the selection of suitable near neighbors with observational noise. In order to evaluate the proposed method, we apply local linear approximation methods to the Henon map corrupted by noise. As a result, we confirm that the noise reduction method works well to improve prediction accuracy.
展开▼
