前反馈神经网络模型在预测本院急诊处方量中的应用效果

Back-propagation network predict the number of emergency prescription

目的探讨前反馈神经网络在本院急诊处方量预测分析中的应用。方法根据本院急诊2012至2018年的处方量及本地区天气情况和流行病发病率情况,采用Excel处理数据后使用R软件建立神经网络模型,并使用2019年处方量数据进行模型验证。结果使用神经网络预测模型实际处方数和预测处方数的趋势一致,均方差(MSE)为19399.51,平均绝对误差百分比(MAPE)为9.13%,能较好的模拟实际处方量。结论神经网络预测模型能较准确预测多种因素影响的非规律性的处方量,有利于降低医院管理成本和提高医疗服务质量。同时,该模型对单品种的适用性及更加友好界面的设计有待进一步讨论。

Objective To discuss predicting the number of emergency prescription by back-propagation network model.Methods According to the prescription volume of emergency department in our hospital from 2012 to 2018 and the weather conditions and epidemic incidence in the region,Excel was used to process the data and R software was used to establish the neural network model,and the prescription volume data in 2019 was used to verify the model.Results Using back-propagation network model,the actual number of prescription and to predict the number of the prescription of consistent,the value of the mean square error(MSE)was 19399.51,the value of the mean absolute percentage error(MAPE)was 9.13%.The results show that model can better simulate the actual number of prescription.Conclusion Back-propagation network model to a variety of factors affect the regularity more accurately predict the quantity of prescription.Can be used to decrease the cost of hospital management and medical service quality,but the model for a single variety of applicability and more friendly interface design needs further discussion.