基于PSO-LSTM算法的医用耗材消耗量预测模型研究

Research on prediction model of medical consumable consumption based on PSO-LSTM algorithm

目的:建立基于粒子群优化长短期记忆(PSO-LSTM)算法的医用耗材消耗量预测模型(PSO-LSTM模型),预测医院医用耗材消耗量,实现医用耗材精细化管理。方法:选取2019年1月至2020年12月医院使用的国家第一批重点监控高值耗材消耗量数据,建立PSO-LSTM预测模型,分析医用耗材消耗情况,预测医用耗材消耗量。采用均方误差(MSE)评价PSOLSTM模型预测医用耗材消耗量数据与LSTM网络数据的误差程度。结果:PSO-LSTM模型预测吻合器消耗量与LSTM网络相比其MSE值降低84%,PSO-LSTM模型预测穿刺器消耗量与LSTM网络相比其MSE值降低77%,PSO-LSTM模型的医用耗材消耗量预测结果与实际消耗量更为接近,医用耗材消耗量预测精度显著高于LSTM网络。结论:PSO-LSTM模型能够准确预测医用耗材消耗量,为医用耗材采购决策提供科学依据,实现医用耗材精细化管理,降低医用耗材使用成本。

Objective:To establish a consumption prediction model of medical consumables based on particle swarm optimization long and short-term memory(PSO-LSTM)algorithm to predict the consumption of medical consumables and achieve fine management of medical consumables.Methods:The consumption data of the first batch of consumption of high-value consumables under key national monitoring used by hospitals from January 2019to December 2020 were selected,the consumption situation was analyzed,and a prediction model for consumption of medical consumables based on PSO-LSTM algorithm based on particle swarm optimization(PSO)and long and short-term memory(LSTM)algorithm was established for prediction of consumption of medical consumables.The mean square error(MSE)was used to evaluate the degree of error of predicted data from actual value.Results:The MSE value of stapler consumption predicted by PSO-LSTM and LSTM was decreased by 84%,and the MSE value of puncture device consumption predicted by PSO-LSTM and LSTM was decreased by 77%.The prediction model of medical consumables consumption based on PSO-LSTM algorithm was closer to the actual consumption,and the prediction accuracy was improved significantly compared with LSTM algorithm.Conclusion:The consumption prediction model of medical consumables based on PSO-LSTM algorithm can accurately predict the consumption of medical consumables,provide suggestions for procurement decisions of medical consumables,finely manage the medical consumables,and reduce the use cost of medical consumables.