基于LSTM时间序列模型的医疗资源配给预测方案

打开文本图片集

中图分类号：TP183 文献标识码：A 文章编号：2096-4706（2025）22-0017-07

Medical Resource Allocation Prediction Scheme Based on LSTM Time Series Model

GUOYing'，ZHANGLili'，LI Yuemei¹，YULiangchao² （1.The First Afliated Hospital ofGuilin Medical University， Guilin 541oo1， China; 2.Guilin University ofElectronic Technology， Guilin 5410o4， China）

Abstract： Accurate prediction and efficient deployment of medical supplies are the core to improve the effectiveness ofepidemicpreventionandcontrol.Thetraditionalmodelisdifculttoefectivelydealwiththenonliearcharacteristicsand dynamic nterferenceof time series data，anditiseasyto introduce prediction bias.BasedontheLong Short-TermMemory （LSTM）network，this studyuses its gating mechanismandlong-term dependence modelingability tooptimizethecomplex temporalcorrelationinepidemicdata.CombinedwithAdamoptimizer，aneffcientmedicalsupplydemandforcastingmodelis constructed.Experimentsbasedontherealdataof thenewcoronavirus epidemicinmultiplecitiesshow thattheLSTMmodel inthis paper has advantages in predictionperformance and eficiencycompared with benchmarks such asregressonmodel， recurentneuralnetworkandTransformer.Theablationexperimentfurthershowsthatwhenthelengthofthehistoricalwindow is setto4，the modelcanbestbalance thelong-termrulecaptureandnoisesuppression，therebyachieving optimalperformance.

Keywords： medical supply demand forecasting; LSTM; time series prediction; gating mechanism

0 引言

近年来，全球范围内公共卫生突发事件 （流感、新冠等）频发，严重威胁人类生命健康[。（剩余10584字）