Abstract

Exponential smoothing is one of the most commonly used prediction methods. When the data has obvious periodicity and seasonality, Holt–Winters usually has a good prediction performance. However, the predicted results often do not meet our expectations when the trend of the original data is not clear. To further reduce the randomness of time series, a new method combining grey generating operator with the traditional Holt–Winters is proposed. The accumulated sequence by grey generating operator can have obvious variation law. Three practical examples were selected to evaluate the forecasting performance of this proposed method. The results indicate that the proposed model can substantially have the better forecasting capability than traditional Holt–Winters method.

摘要

指数平滑是最常用的预测方法之一。当数据具有明显的周期性和季节性时，Holt-Winters 通常具有较好的预测性能。然而，当原始数据的趋势不明确时，预测结果往往不符合我们的预期。为了进一步降低时间序列的随机性，提出了一种将灰色生成算子与传统Holt-Winters相结合的新方法。灰度生成算子的累加序列可以有明显的变化规律。选择了三个实际例子来评估该方法的预测性能。结果表明，所提出的模型比传统的Holt-Winters方法具有更好的预测能力。