基于长短时记忆循环神经网络的北京市糖尿病合并呼吸系统疾病患者入院预测研究

目的：比较广义相加模型（GAM）和长短时记忆循环神经网络（LSTM-RNN）对糖尿病合并呼吸系统疾病患者入院频数的预测效果。方法：收集2014年1月1日至2019年12月31日北京市大气污染物、气象因素和呼吸系统疾病入院数据，基于LSTM-RNN预测糖尿病合并呼吸系统疾病患者入院频数并与GAM对比，模型评价采用五折交叉验证法。结果：与GAM相比，LSTM-RNN具有较低的预测误差[均方根误差（RMSE）分别为21.21±3.30和46.13±7.60，P<0.01；平均绝对误差（MAE）分别为14.64±1.99和36.08±6.20，P<0.01]和较高的拟合优度（R²值分别为0.79±0.06和0.57±0.12，P<0.01）。在性别分层中，预测女性入院频数时，LSTM-RNN三项指标均优于GAM（均P<0.05）；预测男性入院频数时，两模型误差评价指标差异无统计学意义（均P>0.05）。在季节分层中，预测温暖季节的入院频数时，LSTM-RNN的RMSE和MAE均低于GAM（均P<0.05），R²值差异无统计学意义（P>0.05）；预测寒冷季节入院频数时，两种模型的RMSE、MAE和R²值差异均无统计学意义（均P>0.05）。在功能区分层中，预测首都功能核心区入院频数时，LSTM-RNN的RMSE、MAE和R²值均优于GAM（均P<0.05）。结论：LSTM-RNN预测误差较小，拟合程度优，可作为污染天气提前精准配置医疗资源的预测手段。

Research on prediction of daily admissions of respiratory diseases with comorbid diabetes in Beijing based on long short-term memory recurrent neural network

Objective: To compare the performance of generalized additive model (GAM) and long short-term memory recurrent neural network (LSTM-RNN) on the prediction of daily admissions of respiratory diseases with comorbid diabetes. Methods: Daily data on air pollutants, meteorological factors and hospital admissions for respiratory diseases from Jan 1st, 2014 to Dec 31st, 2019 in Beijing were collected. LSTM-RNN was used to predict the daily admissions of respiratory diseases with comorbid diabetes, and the results were compared with those of GAM. The evaluation indexes were calculated by five-fold cross validation. Results: Compared with the GAM, the prediction errors of LSTM-RNN were significantly lower [root mean squared error (RMSE): 21.21±3.30 vs. 46.13±7.60, P<0.01; mean absolute error (MAE): 14.64±1.99 vs. 36.08±6.20,P<0.01], and theR² value was significantly higher (0.79±0.06 vs. 0.57±0.12, P<0.01). In gender stratification, RMSE, MAE andR² values of LSTM-RNN were better than those of GAM in predicting female admission (all P<0.05), but there were no significant difference in predicting male admission between two models (allP>0.05). In seasonal stratification, RMSE and MAE of LSTM-RNN were lower than those of GAM in predicting warm season admission (allP<0.05), but there was no significant difference inR² value (P>0.05). There were no significant difference in RMSE, MAE andR² between the two models in predicting cold season admission (all P>0.05). In the stratification of functional areas, the RMSE, MAE andR² values of LSTM-RNN were better than those of GAM in predicting core area admission (all P<0.05).Conclusion: LSTM-RNN has lower prediction errors and better fitting than the GAM, which can provide scientific basis for precise allocation of medical resources in polluted weather in advance.