Progressive Tuning of Simple Exponential Smoothing Forecasts
提出一种有限样本下的指数平滑方法及平滑参数估计公式，并与渐进数值优化和自适应预测方法在合成与真实数据上比较。
AbstractThe paper outlines a finite sample version of exponential smoothing, and proposes a formula for estimating the smoothing parameter. The resulting method, which can be implemented on a recursive basis over time, is compared with alternative approaches, such as progressive numerical optimization of the smoothing parameter and adaptive forecasting on both synthetic and real data.Keywords: exponential smoothingforecastingKalman filteringtime-series analysis