MEASURING THE PREDICTABILITY OF NOISY RECURRENT TIME SERIES

Autor:	A. Passamante, D. Pickett, R. Wayland, D. Bromley
Rok vydání:	1993
Předmět:	Series (mathematics) Applied Mathematics Noise Linearization Simple (abstract algebra) Modeling and Simulation Prediction methods Statistics Orthogonal least squares Predictability Total least squares Engineering (miscellaneous) Algorithm Mathematics
Zdroj:	International Journal of Bifurcation and Chaos. :797-802
ISSN:	1793-6551 0218-1274
DOI:	10.1142/s0218127493000738
Popis:	The effect of the chosen forecasting method on the measured predictability of a noisy recurrent time series is investigated. Situations where the length of the time series is limited, and where the level of corrupting noise is significant are emphasized. Two simple prediction methods based on explicit nearest-neighbor averages are compared to a more complicated, and computationally expensive, local linearization technique based on the method of total least squares. The comparison is made first for noise-free, and then for noisy time series. It is shown that when working with short time series in high levels of additive noise, the simple prediction schemes perform just as well as the more sophisticated total least squares method.
Databáze:	OpenAIRE
Externí odkaz:	https://explore.openaire.eu/search/publication?articleId=doi_________::d029cdb4e7cbdc0f86e0ba1de4f82665 https://doi.org/10.1142/s0218127493000738 Zobrazit plný text záznamu