基于粒子群优化长短记忆网络算法的有创呼吸机使用量预测模型研究

Research on Prediction Model of Invasive Ventilator Usage Based on Particle Swarm Optimization-Long and Short Memory Network Algorithm

目的研究基于粒子群优化长短记忆网络算法的有创呼吸机使用量预测模型。方法选取2019年4月至2023年4月医院有创呼吸机使用情况数据,建立基于粒子群优化长短记忆网络(PSO-LSTM)算法的有创呼吸机使用量预测模型,预测全院及重症监护病房(ICU)有创呼吸机每天使用数量。采用平均绝对误差(MAE)、平均绝对百分比误差(MAPE)及均方根误差(RMSE)作为准确性评价指标。结果PSO-LSTM模型预测重症ICU有创呼吸机每天在用量与LSTM模型比较,其MAE值降低41.15%、MAPE值降低50%、RMSE值降低44.36%;PSO-LSTM模型预测全院有创呼吸机每天在用量与LSTM模型比较,MAE值降低81.93%、MAPE值降低83.33%、RMSE值降低79.08%,PSO-LSTM模型预测精度高于LSTM模型。结论PSO-LSTM模型能够准确预测有创呼吸机的每天在用量,为有创呼吸机采购决策提供科学依据,为创建全院呼吸机管理共享中心提供数据分析基础,进一步提升医疗设备精细化管理水平。

Objective To study the prediction model of invasive ventilator usage based on particle swarm optimization long and short memory network algorithm.Methods Data on the use of invasive ventilators in the whole hospital were selected from April 2019 to April 2023.A predictive model of invasive ventilator usage based on particle swarm optimization and long and short memory network(PSO-LSTM)algorithm was established to predict the daily use of invasive ventilator in the hospital and intensive care unit.The mean absolute error(MAE),mean absolute percentage error(MAPE)and root mean square error(RMSE)were used for the evaluation accuracy indexes.Results PSO-LSTM model predicted that the MAE value,MAPE value and RMSE value of invasive ventilator in intensive ICU were reduced by 41.15%,50%and 44.36%compared with LSTM network,and PSO-LSTM model predicted that the daily consumption of invasive ventilators in the hospital was reduced by 81.93%,83.33%,and 79.08%compared with LSTM network.The PSO-LSTM of the proposed method is significantly higher than that of the LSTM network.Conclusion The PSO-LSTM model can accurately predict the daily consumption of invasive ventilators,provide a scientific basis for the procurement decision of invasive ventilators,provide a data analysis basis for the establishment of a ventilator management sharing center in the whole hospital,and improve the refined management of medical equipment.