【摘要】 目的 分析2014―2021年福建省道路交通伤害死亡情况，探索适用的趋势预测模型。方法 采用季节差分自回归移动平均(seasonal autoregressive integrated moving average, SARIMA)、支持向量回归(support vector regression, SVR)、长短期记忆(long short-term memory, LSTM)网络拟合2014年1月―2021年6月福建省道路交通伤害死亡率数据，对2021年7―12月死亡率进行预测，并与实际值验证比较，评价模型的预测效果。结果 2014―2021年福建省道路交通伤害年报告率呈下降趋势(AAPC=-6.29%,P<0.001)。模型比较LSTM网络预测准确度最好，均方根误差(root mean square error, RMSE)、平均绝对误差(mean absolute error, MAE)和平均绝对百分比误差(mean absolute percentage error, MAPE)分别为0.070 5、0.061 2、8.72%。结论 福建省道路交通伤害死亡总体呈下降趋势，LSTM网络可用于道路交通伤害死亡率的短期预测。更多还原

【Abstract】 Objective To analyze the status of road traffic injury deaths in Fujian province from 2014 to 2021,and to explore the applicable trend prediction model. Methods Seasonal autoregressive integrated moving average(SARIMA),support vector regression(SVR) and long short-term memory(LSTM) network were constructed using road traffic injury deaths data from January 2014 to June 2021 in Fujian province to predict the mortality rate from July to December in 2021, and its prediction effects were evaluated by comparison with the actual value. Results The annual reporting rate of road traffic injuies in Fujian Province showed a decreasing trand from 2014 to 2021(AAPC=-6.29%, P<0.001). LSTM network had highest prediction accuracy among three models, with root mean square error(RMSE) of 0.070 5, mean absolute error(MAE) of 0.061 2 and mean absolute percentage error(MAPE) of 8.72%. Conclusions The overall road traffic injury deaths in Fujian province showed a downward trend. LSTM network can be used to predict the short-term trend of road traffic injury deaths.更多还原