摘要
针对新型传染病发展趋势的预测精度问题,提出一种改进粒子群(PSO)算法优化卷积神经网络(CNN)与长短期记忆神经网络(LSTM)相结合的预测模型.首先,将原始粒子群优化算法中最优惯性权重的调整方式由迭代次数的线性关系转变为非线性关系,并对学习因子进行线性更新,以寻找最优参数,从而更准确地模拟粒子群的社会学习能力,进而平衡算法的全局优化能力,提高收敛速度;其次,以发酵时间较长的新型冠状肺炎为研究对象,构建CNN-LSTM神经网络预测模型,利用CNN层提取其特征信息后降维作为LSTM层输入,并通过预测模块实现对研究对象的指标训练和预测,从而提高模型的预测精度;最后,与原始LSTM模型的预测误差,如均方根误差(RMSE)、平均绝对误差(MAE)、均方误差(MSE)等指标进行对比.研究结果表明,在训练集上,与原始LSTM模型相比,经过改进的PSO算法优化CNN-LSTM组合神经网络模型,其在RMSE、MAE和MSE三个指标上分别降低了73.0%、62.3%、92.7%;在测试集上,这3个指标分别降低了23.0%、29.8%、40.7%.这说明该模型具有更小的误差和较好的预测效果.该研究结果可为实现传染病传播趋势的精准预测提供新的思路和方法.
Aiming at the prediction accuracy of the development trend of new infectious diseases,an improved particle swarm optimization(PSO)algorithm is proposed to optimize the prediction model combining convolutional neural network(CNN)and short term memory neural network(LSTM).First,the adjustment method of the optimal inertia weight in the original Particle Swarm Optimization algorithm is transformed from a linear to a nonlinear relationship by the number of iterations,and the learning factors are linearly updated to find the optimal parameters which can more accurately simulate the social learning ability of the particle swarm,thus it can balance the global optimization ability of the algorithm and improve the convergence rate.Second,a CNN-LSTM neural network prediction model is established with a novel coronary pneumonia with a long fermentation time as the research object,and the CNN layer is used to extract its feature information and then downscaled as the input of the LSTM layer,and the prediction module is used to realize the index training and prediction of the research object,so as to improve the model prediction accuracy.Finally,the predictions obtained from the original LSTM model,and were compared with the following three indices:root mean square error(RMSE),mean absolute error(MAE),and mean square error(MSE).Our results show that the improved PSO algorithm optimized CNN-LSTM combined neural network model reduces the three indicators by 73.0%,62.3%,and 92.7%compared to the original LSTM model on the training set;On the test set,the three indicators decreased by 23.0%,29.8%,40.7%,respectively,it shows that the model has smaller error as well as better prediction effect,which provides ideas and methods to achieve accurate prediction of infectious disease transmission trend.
作者
刘彩云
聂伟
孟金葆
张涛
LIU Caiyun;NIE Wei;MENG Jinbao;ZHANG Tao(School of Mathematics and Computer,Tongling University,Tongling 244061,China;Department of Information Engineering,Tongling Polytechnic,Tongling 244061,China)
出处
《湖州师范学院学报》
2024年第4期37-48,共12页
Journal of Huzhou University
基金
安徽省高校基金资助项目(2023AH040229,2022jyxm1646)
铜陵学院大学生科研项目(2021tlxydxs121)
铜陵职业技术学院科研重点项目(TZY23ZRZD01).
关键词
长短期记忆神经网络
卷积神经网络
粒子群算法
传染病预测
long short term memory neural networks
convolutional neural networks
particle swarm algorithm
infectious disease forecast
