四层参数自调整BP神经网络模型及其在人口死亡率预测中的应用

相鑫,刘秀丽

系统科学与数学 ›› 2018, Vol. 38 ›› Issue (6) : 702-710.

PDF(598 KB)
PDF(598 KB)
系统科学与数学 ›› 2018, Vol. 38 ›› Issue (6) : 702-710. DOI: 10.12341/jssms13412
论文

四层参数自调整BP神经网络模型及其在人口死亡率预测中的应用

    相鑫1,3,刘秀丽1,2,3
作者信息 +

Four-Layer Parameter Self-Adjusting BP Neural Network Model and Its Application in Population Mortality Prediction

    XIANG Xin 1,3 ,LIU Xiuli 1,2,3
Author information +
文章历史 +

摘要

为有效提高神经网络的学习效率并降低其陷入局部极小值的概率, 文章构建了四层参数自调整BP神经网络模型, 该模型具有四层特殊的网络结构, 采用附加动量法与自适应法调整参数. 数值试验表明, 与常规的BP神经网络相比, 该方法的学习收敛速度与预测精度均有较大改进. 鉴于人口死亡率的复杂变化趋势, 使用单个模型无法准确预测, 文章同时建立了GM(1,1)模型与ARMA(2,2)模型, 结合三种模型的优势, 应用模型平均方法预测了中国人口的死亡率. 结果显示, 2018--2020 年中国人口的死亡率分别为7.1042\textperthousand, 7.1040\textperthousand ~和7.1045\textperthousand.

Abstract

In order to effectively improve the learning efficiency of the neural network and reduce its probability of being trapped in the local minimum, a four-layer parameter self-adjusting BP neural network model was constructed in this paper. The model had four special network structures, used additional momentum method and adaptive method to adjust parameters. Numerical experiments showed that compared with the conventional BP neural network, the learning convergence speed and prediction accuracy of this method were greatly improved. Due to the complex changes in the mortality rate of the population, using a single model can not be accurately predicted. This paper also established the GM(1,1) and the ARMA(2,2) model, combined the advantages of these models, used the model average method to predict the mortality rate of our population. The results showed that the mortality rate of the population in China during 2018--2020 would be 7.1042\textperthousand, 7.1040\textperthousand ~and 7.1045\textperthousand, respectively.

关键词

BP神经网络 / 模型平均 / 人口死亡率 / 预测.

引用本文

导出引用
相鑫 , 刘秀丽. 四层参数自调整BP神经网络模型及其在人口死亡率预测中的应用. 系统科学与数学, 2018, 38(6): 702-710. https://doi.org/10.12341/jssms13412
XIANG Xin , LIU Xiuli. Four-Layer Parameter Self-Adjusting BP Neural Network Model and Its Application in Population Mortality Prediction. Journal of Systems Science and Mathematical Sciences, 2018, 38(6): 702-710 https://doi.org/10.12341/jssms13412
PDF(598 KB)

Accesses

Citation

Detail

段落导航
相关文章

/