search query: @journal_id 525 / total: 244
reference: 2 / 244
Author: | Cipra, T. |
Title: | Robust exponential smoothing |
Journal: | Journal of Forecasting
1992 : JAN, VOL. 11:1, p.57-69 |
Index terms: | STATISTICAL METHODS TIME SERIES FORECASTING FREQUENCY DISTRIBUTION ALGORITHMS |
Language: | eng |
Abstract: | Robust modifications of exponential smoothing for time series with outliers or long-tailed distributions are considered. Classical smoothing applied to such time series is sensitive to the presence of outliers or long-tailed distributions and may give inadequate smoothing and forecasting results. Simple and double exponential smoothing in the L1 norm are discussed, then general exponential smoothing is made robust, replacing the least squares approach by M-estimation so that the recursive character or the final formula be preserved. Simple algorithmic procedures are given that preserve the advantageous features of exponential smoothing and that are less sensitive to outliers. |
SCIMA