Zhang Shanhua, An Hong Ki, Yin Hongmei
Department of Digital Equipment, Jiangsu Vocational College of Electronics and Information, Huai'an, 223003, China.
Faculty of Civil Engineering & Technology, Universiti Malaysia Perlis, Arau Perlis, 02600, Malaysia.
Heliyon. 2024 Aug 8;10(16):e35889. doi: 10.1016/j.heliyon.2024.e35889. eCollection 2024 Aug 30.
The GM(1,1) model's prediction accuracy is significantly influenced by the accuracy of background value estimation. The traditional trapezoidal background value can only be applied to a specific data sequence. Therefore, this study proposes a GM(1,1) model background value reconstruction approach based on the combination of intelligent trapezoidal and variable weights in order to increase the model's application as well as its prediction accuracy. The trapezoidal background value function with slope and point position parameters is called model I. Then, a set of point position parameter sequences, with a new background value function is constructed, called model II. A genetic algorithm is utilized to seek for the values of the parameters to be determined in both models I and II. The results showed that for the exponential growth data series, model I and II have higher prediction accuracy compared to traditional models. For data sequences, taking the traffic volume series of a road from 2014 to 2023, the prediction accuracy of this paper's model I method can be improved by 0.3643 % and 0.2725 % compared with Deng's and Wang's models. The prediction accuracy of this paper's model II method has been further improved by 0.1075 % compared with that of model I.
GM(1,1)模型的预测精度受背景值估计精度的显著影响。传统梯形背景值仅适用于特定的数据序列。因此，本研究提出一种基于智能梯形和变权组合的GM(1,1)模型背景值重构方法，以扩大模型的应用范围并提高其预测精度。具有斜率和点位置参数的梯形背景值函数称为模型I。然后，构建一组点位置参数序列，得到一个新的背景值函数，称为模型II。利用遗传算法求解模型I和模型II中待确定参数的值。结果表明，对于指数增长数据序列，模型I和模型II比传统模型具有更高的预测精度。对于数据序列，以某道路2014年至2023年的交通量序列为例，本文模型I方法的预测精度与邓氏模型和王氏模型相比，分别提高了0.3643%和0.2725%。本文模型II方法的预测精度与模型I相比进一步提高了0.1075%。
