The new method of measuring the effects of noise reduction in chaotic data

The presence of a noise, which is typical for real data, makes methods of chaotic signals analysis much more difficult to apply to. That is why algorithms of noise reduction in chaotic time series have been recently developed. A lot of existing algorithms require setting values of specified parameters and in consequence lead to many outputs. Thus one must additionally apply a supporting method which allows to indicate a “proper” output. In this paper such a new method is proposed and examined. As an example, the presented method is applied to support the Nearest Neighbours algorithm to reduce the noise in the time series from the Warsaw Stock Exchange. Next the cleaned data are investigated for the presence of chaos.