四川大学学报（自然科学版）2025，Vol.62Issue(3)：745-750,6.DOI:10.19907/j.0490-6756.240352
一种改进的差分-指数平滑预测模型
An improved differential-exponential smoothing forecasting model
摘要
Abstract
Differential-exponential smoothing model(DESM)has been extensively used in the prediction for time series.The model requires first performing stationary differencing on the time series,and then applying exponential smoothing techniques for forecasting.When there exist significant fluctuations in time series,the obtained time-series may still be non-stationary even the second-order differencing is performed.In this case,the predictive performance of DESM may be seriously damaged.To overcome this issue,we propose the im-proved DESM(IDESM)by adjusting the non-stationary data in the second-order differenced time series of the model.The predictive performances of IDESM,DESM and the residual auto-regressive model(RARM)are compared.It is shown that the prediction accuracy of IDESM is significantly better than the latter two models.Moreover,IDESM has good predictive performance for the time series containing different growth trends.关键词
差分-指数平滑模型/残差自回归模型/时间序列/预测Key words
Difference-exponential smoothing model/Residual auto-regressive model/Time series/Fore-casting分类
数理科学引用本文复制引用
秦茜茜,王圆秋..一种改进的差分-指数平滑预测模型[J].四川大学学报（自然科学版）,2025,62(3):745-750,6.基金项目
国家自然科学基金(12075162) （12075162）
中央引导地方科技发展项目(2024ZYD0115) （2024ZYD0115）
四川省自然科学基金(2024NSFSC0040) （2024NSFSC0040）
