基于LSTM深度神经网络的月门诊量预测精度研究  

The Accuracy of Monthly Outpatient Volume Prediction Based on LSTM Deep Neural Network

作　　者：李琳[1] 王哲 张学良[1] 王凯[1] 周毅[2] LI Lin;WANG Zhe;ZHANG Xue-liang(Computer Center of Medical College,Sun Yat-sen University,Guangzhou 510080,Guangdong Province,P.R.C)

机构地区：[1]新疆医科大学 [2]中山大学中山医学院计算机中心

出　　处：《中国数字医学》2019年第1期14-17,共4页China Digital Medicine

基　　金：国家自然科学基金项目(编号:61876194;11661007);国家重点研发计划项目(编号:2018YFC0116902;2018YFC0116904;2016YFC0901602);NSFC-广东大数据科学中心联合基金项目(编号:U1611261);广州市科技计划项目(编号:201604020016)~~

摘　　要：目的:分析新疆地区慢性阻塞性肺病的月门诊量变化趋势,对医院月门诊量预测方法进行探讨,为医院合理配置医疗资源和提高救助能力提供科学依据。方法:采用ARIMA模型和LSTM模型对新疆地区慢性阻塞性肺病的月门诊量的时间序列进行预测,使用RMSPE值评价不同方法的预测精度。结果:ARIMA模型、时间步为1的LSTM、时间步为12的LSTM的RMSPE值分别是20.23%、22.23%和20.01%,相较之下时间步为12的LSTM网络的预测效果较好,时间步为1的LSTM效果最差。结论:LSTM预测医院月门诊量的准确率较高,为医院月门诊量预测提供了新的方法。Objective: The variation trend of monthly outpatient volume of chronic obstructive pulmonary disease(COPD) in XinJiang was analyzed. The forecasting method of monthly outpatient volume of hospital was discussed to provide scientific basis for rational allocation of medical resources and improvement of rescue capability of hospital. Methods: ARIMA model and LSTM model were used to predict the time series of monthly outpatient volume of COPD in XinJiang and RMSPE value was used to evaluate the prediction accuracy. Results: RMSPE values of ARIMA models, LSTM network with time step 1, LSTM network with time step 12 were 20.23%, 22.23% and 20.01% respectively, compared with the predicted effect of LSTM network with time step 12, and the effect of LSTM with time step 1 was the worst. Conclusion: The accuracy of LSTM in predicting hospital outpatient volume is higher, which provides a new method for predicting hospital outpatient volume.

关 键 词：LSTM 门诊量 ARIMA 预测 慢性阻塞性肺病 

分 类 号：R319[医药卫生—基础医学]