摘要
目的:构建医用耗材需求预测模型预测医用耗材需求,减少医用耗材紧急补货和缺货事件发生.方法:运用支持向量回归(SVR)算法建立医用耗材需求预测模型,采用多种群遗传算法(MPGA)建立补货策略和库存控制策略寻优模型构建医用耗材需求预测模型,自动生成补货策略和库存控制策略.选取2019年1月至2023年5月本溪市中心医院引入供应-加工-配送(SPD)模式进行医用耗材管理的相关数据,分别用于构建模型训练、模型验证测试(测试集)和模型应用预测(应用集).评估模型的平均绝对预测误差值、预测误差峰值和预测误差谷值,对比模型应用的周日均消耗与库存成本比值、月均紧急订货次数、月均缺货次数、月非紧急订货次数和指标周平均值降低幅度.结果:模型测试集和应用集的平均绝对预测误差值比较差异无统计学意义(P＞0.05);模型平均绝对预测误差值为(0.033 5±0.024 5),预测误差峰值为0.071 7,预测误差谷值为-0.009 0.模型应用后周日均消耗与库存成本比值、月均紧急订货次数、月均缺货次数和月非紧急订货次数分别为(0.457 5±0.060 3)、(23.95±6.04)次、(5.58±2.17)次和(20.68±2.77)次,周日均消耗与库存成本比值高于应用前,而月均紧急订货次数、月均缺货次数和月非紧急订货次数均低于应用前,差异均有统计学意义(F=371.912、88.486、124.472、142.138,P＜0.000);模型应用后周平均库存金额降低43.66%,平均紧急补货次数降低53.76%,平均缺货次数降低76.95%,平均正常补货次数降低34.41%.结论:医用耗材需求预测模型能够预测医用耗材需求,优化补货策略和库存控制策略,降低医用耗材库存成本,减少紧急补货和缺货现象发生,并可减少正常补货次数.
Abstract
Objective:To build a medical consumables demand prediction model to predict the demand for medical consumables,and to reduce the occurrence of emergency replenishment and shortage of medical consumables.Methods:The medical consumables demand prediction model was established using support vector regression(SVR)algorithm,the replenishment strategy and inventory control strategy optimization model was established using multiple population genetic algorithm(MPGA)to build a medical consumables demand forecasting model,and the replenishment strategy and inventory control strategy were automatically generated.The relevant data of supply-processing-distribution(SPD)mode introduced by Benxi Central Hospital for medical consumables management from January 2019 to May 2023 were selected for model training,model validation test(test set)and model application prediction(application set),respectively.The average absolute prediction error,peak prediction error and trough of prediction error of the model were evaluated,and the average ratio of consumption per day to inventory cost,the average number of urgent orders per month,the average number of out of stock per month,the average number of non-urgent orders per month and the decreasing range of the index weekly average were compared.Results:There was no statistically significant difference in the average absolute prediction error between the model test set and the application set(P＞0.05).The average absolute prediction error of the model was 0.033 5±0.024 5,the peak prediction error was 0.071 7 and the trough of prediction error was-0.0090.After the application of the model,the ratio of average daily consumption to inventory cost,the average number of monthly emergency order frequency,the average number of shortage of stock,and the average number of monthly non-emergency orders were(0.457 5±0.060 3),(23.95±6.04),(5.58±2.17),and(20.68±2.77),respectively.The ratio of daily average consumption to inventory cost was higher than that before the application,the average number of monthly emergency order,shortage of stock,and monthly non-emergency order were lower than those before the application,the difference was statistically significant(F=371.912,88.486,124.472,142.138,P＜0.000).After the application of the model,the weekly average inventory amount decreased by 43.66%,and the average number of emergency replenishment decreased by 53.76%,the average number of shortages of stock decreased by 76.95%,and the average shortage of stock normal replenishments decreased by 34.41%.Conclusion:The medical consumables demand prediction model can predict the demand for medical consumables,optimize replenishment and inventory control strategies,reduce the cost of medical consumables inventory,reduce occurrence of emergency replenishment and shortages of stock,and reduce the number of normal replenishment.
国家科技期刊平台
NSTL
万方数据