一种对二次指数平滑预测中参数和初值的研究

Study on choosing coefficient of secondary exponential smoothing and initial value

针对指数平滑法在实际应用主要靠经验选取平滑参数和初值的不足,通过对传统预测模型的分析,提出了递推分割法来优选平滑参数以及初值选取模型化的思路,得到了比较优秀的平滑参数,获取了以预测的实际值n为划分依据的简便方法处理后的初值。然后将该优化方法应用于股票"三连商社"的预测,取得了明显优于传统模型的结果。

Due to the deficiencies of the exponential smoothing method that coefficient of exponential smoothing and initial values are chosen mainly relying on experiences. Through the analysis of traditional forecast model, the thought of choosing coefficient of exponential smoothing and the initial value by progression dichotomous search （PDS） is proposed. Therefore, a more outstanding coefficient of exponential smoothing is obtained while the initial value is acquired by the simple methods divided on the basis of forecasted practical value n. Moreover, this new model is applied to the forecast of the stock of ＂Sanlian Commerce Co., Ltd＂ which proves obvious superiority to traditional model.