Fig. 6.

The system tracking performance of the adaptive chaos synchronization method: a noiseless data sampled at 1 min intervals b data with 20% proportional error sampled at 1 min intervals; c data with 50% proportional error sampled at 1 min intervals; d noiseless sparse data sampled at 45 min intervals. The data were filtered by wavelet denoising. Predictions based on chaos synchronization and grid search track the input data with relatively high accuracy. The adaptive chaos synchronization algorithm combined with grid search avoids local minima even when the choice of initial values of the parameters is poor. Although we display results corresponding to one cycle for clear visualization, the prediction closely overlap the input data for the entire six-cycle data set