摘要
目的探讨极限梯度提升(extreme gradient boosting,XGBoost)和长短期记忆网络(long short-term memory,LSTM)算法模型对区域救护车需求量的预测价值,从而选择最优的区域救护车需求量预测模型。方法收集2009~2018年某大学附属医院共40014条救护车呼叫记录,以及3968条天气数据,构建XGBoost和LSTM预测模型,用平均绝对百分比误差(mean absolute percentage error,MAPE)和平均绝对误差(mean absolute error,MAE)来比较两种模型预测准确性。结果XGBoost模型和LSTM模型中预测每日救护车需求量的MAPE值分别为24.29%和17.47%,MAE值分别为2.692和2.462,LSTM模型预测的准确性优于XGBoost模型。结论XGBoost和LSTM网络预测模型对区域救护车需求量均有较好的预测价值,其中LSTM模型预测效果更优。
Objective To study the value of the forecast models by machine learning[extreme gradient boosting(XGBoost)and long short-term memory(LSTM)model]for the daily volume of regional ambulance using,and to select the best forecast model.Methods A total of 40014 ambulance call records from the Second Affiliated Hospital of Guangzhou Medical University and 3968 weather data from 2009 to 2018 were collected.XGBoost and LSTM forecast models were constructed by machine learning methods.Mean absolute percentage error(MAPE)and mean absolute error(MAE)were used to assess the accuracy of two forecast models.Results The MAPE values of XGBoost model and LSTM model for daily ambulance demand were 24.29%and 17.47%,respectively.The MAE values were 2.692 and 2.462 respectively.Therefore,the accuracy of the LSTM forecast model was better than that of the XGBoost model.Conclusions Both XGBoost and LSTM models have good predictive value for the regional ambulance demand.The predictive value of LSTM model is better.
作者
刘佳
江慧琳
王静
伍卓文
李双明
曾睿
罗一洲
黄海铨
茅海峰
程琦
伍宝玲
陈晓辉
Liu Jia;Jiang Hui-lin;Wang Jing;Wu Zhuo-wen;Li Shang-ming;Zeng Rui;Luo Yi-zhou;Huang Hai-quan;Mao Hai-feng;Cheng Qi;Wu Bao-ling;Chen Xiao-hui(School of Public Health,Guangzhou Medical University,Guangzhou 511436,China)
出处
《中国急救医学》
CAS
CSCD
2022年第10期893-898,共6页
Chinese Journal of Critical Care Medicine
作者简介
刘佳(1997-),女,硕士研究生,E-mail:978462001@qq.com;通信作者:陈晓辉(1965-),男,教授,E-mail:cxhgz168@126.com。
