一种优选移动平均预测模型的Min-Min算法

为使移动平均法预测技术的应用从专业化向大众化、人工化向智能化转变,在对一次移动平均模型进行改进之后,以预测的局部残差平方和最小为原则,设计出一种优选移动平均预测模型的Min-Min算法:首先,分别选出一次移动平均模型和二次移动平均模型各自的最优移动项数;然后,在最优的一次模型与最优的二次模型之间作出最优次数的选择;最后,基于优选出的移动平均模型对未来一期开展点预测和区间预测.同既有算法相比,本算法对移动平均法预测技术的进步性主要体现在:a)以先选定移动项数、后选定移动次数的程序算法取代先选定移动次数、后选定移动项数的专家做法,从而将移动平均法的实施从半自动化的人工预测提升至全自动化的智能预测;b)对现行的一次移动平均模型进行改进,从而大幅提高一次移动平均法的预测能力;c)在移动平均模型现行只有点预测的基础上进一步提出区间预测,从而起到完善和丰富预测报告的作用.

Min-min algorithm for optimizing moving average prediction model

After improvement of one-time moving average model, this paper designed a Min-Min algorithm to optimize the moving average prediction model based on the principle of the least summation of the local residual squares for making the application of moving average prediction technology change from specialization to popularization, from artificial to intelligent. Firstly, it respectively selected the optimal number of moving items of the one-time moving average model and the double-time moving average model. Then, it selected the optimal number of moving times between the optimal one-time moving average model and the optimal two-time moving average model. Finally, based on the optimized moving average model it carried out the point prediction and interval prediction for the future first period. Compared with the existing algorithms, the progress of this moving average method was mainly reflected in: a) Replaced expert practice, which first selected the number of moving items and then the number of moving times, by program algorithm, which first selected the number of moving times and then the number of moving items, so as to improve the implementation of moving average method from semi-automatic artificial prediction to full-automatic intelligent prediction. b) Improved the current one-time moving average model, so as to greatly improve the prediction ability of one-time moving average method. c) Further put forward interval prediction model based on the point prediction model, so as to improve and enrich the prediction report of moving average prediction method.