摘要
为了精确预测非线性、非平稳性的血糖浓度时间序列,文章提出一种基于变分模态分解和粒子群优化长短时记忆网络的血糖浓度短期预测模型(VMD-PSO-LSTM)。该方法首先利用VMD方法将患者的血糖浓度时间序列进行分解,得到不同频段的固有模态分量(IMF),然后对各个血糖IMF分量采用经PSO优化的LSTM网络建立预测模型,得到每个IMF分量的预测结果;最后,对各个IMF分量的预测结果进行累加,得到血糖浓度时间序列的最终预测值。基于实测血糖数据进行验证,相较于PSO-LSTM和VMD-LSTM方法,所提方法的平均绝对误差(MAE)、均方根误差(RMSE)以及克拉克网格误差(CEGA)都最小,表明文章提出的VMD-PSO-LSTM模型具有更高的预测精度。
In order to accurately predict the nonlinear and non-stationary blood glucose concentration time series,this paper proposes a short-term prediction model of blood glucose concentration(VMD-PSO-LSTM)based on variational modal decomposi-tion and particle swarm optimization long-term memory network.This method first uses the VMD method to decompose the patient's blood glucose concentration time series to obtain the intrinsic modal components(IMF)of different frequency bands,and then uses the PSO optimized LSTM network to establish a prediction model for each blood glucose IMF component.Finally,the prediction re-sults of each IMF component are accumulated to obtain the final predicted value of the blood glucose concentration time series.It is verified based on measured blood glucose data.Compared with the PSO-LSTM and VMD-LSTM methods,the method proposed in this paper has the smallest mean absolute error(MAE),root mean square error(RMSE)and clarke error grid analysis(CEGA),in-dicating the VMD-PSO-LSTM model proposed in this paper has higher prediction accuracy.
作者
童梦
丁国荣
余楠
王文波
TONG Meng;DING Guorong;YU Nan;WANG Wenbo(College of Science,Wuhan University of Science and Technology,Wuhan 430065)
出处
《计算机与数字工程》
2023年第6期1439-1443,1449,共6页
Computer & Digital Engineering
