基于ARIMA - SVM组合模型的道路交通伤害死亡率预测

目的 探讨ARIMA - SVM组合模型在道路交通伤害死亡率预测中的应用，并与单纯ARIMA模型的预测效果比较。方法 利用2014年1月－2018年6月河北省道路交通伤害死亡率数据拟合建立ARIMA模型和ARIMA - SVM组合模型，对2018年7－12月死亡率进行预测，并与实际死亡率进行验证比较，评价模型的预测效果。结果 ARIMA模型预测值与实际值的相对误差在0.00%～13.38%之间，ARIMA - SVM组合模型预测值与实际值的相对误差在0.00%～3.75%之间；且两者预测效果评价指标RMSE、MAE和MAPE分别为0.102、0.042，0.079、0.036，5.264%、2.469%。结论 ARIMA - SVM组合模型的预测效果优于单纯ARIMA模型，预测精度更高，可用于道路交通伤害死亡率的预测。

Mortality prediction of road traffic injury based on ARIMA-SVM combination model

Objective To explore the application of autoregressive integrated moving average model(ARIMA-SVM) combination model to the prediction of road traffic injury mortality and compare the prediction effect with that of ARIMA model alone.Methods An ARIMA model and a combined ARIMA-SVM model were fitted using road traffic injury mortality data from January 2014 to June 2018 in Hebei Province to predict the mortality rate from July to December in 2018, and the prediction effects of the models were evaluated by comparison with the actual mortality rate. Results The relative errors between the predicted and actual values of the ARIMA model ranged from 0.00% to 13.38%, and the relative errors between the predicted and actual values of the combined ARIMA-SVM model ranged from 0.00% to 3.75%. The evaluation indexes of the prediction effects of both models, root mean square error(RMSE), mean absolute error(MAE) and mean absolute percentage error(MAPE) were 0.102 and 0.042, 0.079 and 0.036, 5.264% and 2.469%. Conclusion The combined ARIMASVM combination model has better prediction effect and higher prediction accuracy than the simple ARIMA model for road traffic injury mortality prediction.