摘要
目的基于双向长短期记忆神经网络(bidirectional long short-term memory,BiLSTM)建立临床用血需求预测模型,预测未来长短期的血液需求量。方法以2017—2022年费县人民医院输血科的临床用血数据为基础建立模型,利用多变量链式方程补全法(multivariate imputation by chained equations,MICE)进行缺失数据补全,将补全后的数据归一化后获得6个血液指标的整体变化趋势。在模型的搭建中,首先,将2017年1月—2021年1月的临床用血数据作为训练集,建立BiLSTM模型;然后,将2021年2月—2022年1月的临床用血数据作为测试集,并获得6个血液指标的预测结果。采用均方根误差(root-mean-square error,RMSE)及平均绝对误差(mean absolute error,MAE)衡量预测精度。结果与传统的LSTM模型相比,BiLSTM模型训练集损失函数的下降更平稳,且平稳后的训练集和测试集的损失函数分别下降了约0.01与0.02。同时,BiLSTM预测的6个指标的平均RMSE和MAE分别为74.18和71.54,相比较LSTM分别下降了33.027%和16.794%。结论BiLSTM可用于临床用血的长短期预测,为中心血站中血制品的制备和调度提供参考。
Objective To establish a clinical blood demand prediction model based on Bidirectional Long Short-term Memory(BiLSTM)neural network and predict future blood demand in long and short term.Methods The paper established a model based on clinical blood usage data from Blood Transfusion Department of Feixian County People's Hospital from 2017 to 2022.Missing data were completed with multivariate input by chain equations(MICE)method,completed data were normalized to achieve overall trend of six blood indicators.During construction of the model,clinical blood data from January 2017 to January 2021 were treated as training set to establish the BiLSTM model.Then,clinical blood data from February 2021 to January 2022 were treated as test set to achieve predicted results.Prediction accuracy was measured with root mean squared error(RMSE)and mean absolute error(MAE).Results Decrease in training set loss function of the BiLSTM model was more stable than traditional LSTM model,and loss functions of stable training set and test set decreased by about 0.01 and 0.02 respectively.Meanwhile,average RMSE and MAE of six indicators predicted by BiLSTM were 74.18 and 71.54 respectively,and decrease of LSTM was 33.027%and 16.794%by comparison.Conclusion BiLSTM can predict short and long term of clinical blood use,and provide reference for preparation and scheduling of blood products in central blood stations.
作者
王全红
祝晓倩
WANG Quanhong;ZHU Xiaoqian(Blood Transfusion Department,Feixian County People's Hospital,Linyi,Shandong 273400)
出处
《智慧健康》
2024年第22期14-19,共6页
Smart Healthcare
