摘要：Exponential smoothing methods forecast time series by discounted past observations. They have become very popular because of their (relative) simplicity compared to their good overall performance. Common applications range from business tasks (e.g., forecasting of sales or stock fluctuations) to environmental studies (e.g., measurements of atmospheric components or rainfall data)—with typically no more a priori knowledge than the possible existence of trend of seasonal patterns. Such methods are sometimes also called naive because no covariates are used in the models, i.e., the data are assumed to be self-explaining. Their success is rooted in the fact that they belong to a class of local models which automatically adapt their parameters to the data during the estimation procedure and therefore implicitly account for (slow) structural changes in the training data. Moreover, because the influence of new data is controlled by hyperparameters, the effect is a smoothing of the original time series.
