基于改进塘鹅算法优化BP神经网络的新型冠状病毒疫情预测  

Prediction of COVID-19 Epidemic Situation by Optimized BP Neural Network Based on Improved Gannet Optimization Algorithm

作　　者：祁慧玲 胡红萍[1] 白艳萍[1] 侯强[1] QI Huiling;HU Hongping;BAI Yanping;HOU Qiang(School of Mathematics,North University of China,Taiyuan 030051,China)

机构地区：[1]中北大学数学学院,山西太原030051

出　　处：《山西大学学报（自然科学版）》2023年第6期1283-1292,共10页Journal of Shanxi University(Natural Science Edition)

基　　金：山西省基础研究计划(20210302123019,20210302123031,202103021224195);山西省回国留学人员科研项目(2020-104)。

摘　　要：针对北京市和美国新型冠状病毒感染疫情的累计确诊人数的预测,提出了基于改进塘鹅优化算法(Improved Gannet Optimization Algorithm,IGOA)、优化反向传播神经网络(Back Propagation Neural Network,BP)权值和偏差的预测模型。将原塘鹅优化算法中塘鹅的位置更新方式由迭代次数的线性关系修改为非线性关系,并引入了随机数,更准确地模拟塘鹅的捕食过程;在两个阶段均采用随机选择机制,并改进了开发阶段塘鹅的位置更新方式,有效地平衡了探索阶段和开发阶段。在实验阶段利用23个基准函数的极值寻优验证了IGOA的有效性,并建立预测模型IGOA-BP,对北京市和美国新型冠状病毒感染累计确诊人数进行预测。与其他7种比较模型相比,预测模型IGOA-BP的预测结果的均方误差(Mean Square Error,MSE)、平均绝对百分比误差(Mean Absolute Percentage Error,MAPE)、平均绝对误差(Mean Absolute Error,MAE)、均方根误差(Root Mean Square Error,RMSE)均最小,表明预测模型IGOA-BP的预测效果最好,对疫情防控政策的制定有一定的参考意义。Aiming at the prediction of the cumulative number of confirmed cases of Corona Virus Disease 2019 in Beijing and America,a prediction model by optimizing the weights and deviations of Back Propagation Neural Network(BP)neural network based on Improved Gannet Optimization Algorithm(IGOA)was proposed.The position updating method of gannet in the original Gannet Optimization Algorithm was changed from the linear relationship of iteration number to the nonlinear relationship,and the random number was introduced to accurately simulate the hunting process of gannet.The random selection mechanism was adopted in both stages,and the location update mode of the gannet in the development stage was improved,which effectively balanced the exploration stage and the development stage.In the experimental stage,the effectiveness of IGOA was verified by the extremum optimization of 23 benchmark functions,and the prediction model IGOA-BP was established to predict the cumulative number of confirmed cases of new coronary pneumonia in Beijing and the United States.The experimental results showed that compared with the other seven comparison models,the MSE,MAPE,MAE and RMSE of IGOA-BP prediction model were the smallest.So the prediction effect of the prediction model IGOA-BP was the best,which could provide good reference for epidemic prevention and control policies.

关 键 词：改进塘鹅优化算法 BP神经网络 新型冠状病毒感染 疫情预测 

分 类 号：O436[机械工程—光学工程]