论文部分内容阅读
目的基于自回归移动平均模型(autoregressive integrated moving average model,ARIMA)-广义回归神经网络模型(generalized regression neural network,GRNN)联合构建交通伤死亡预测模型,并据此对相关公共政策对交通伤死亡的影响进行初步分析。方法 1991~2009年度交通伤死亡人数来源于中国公安部公布的相应年统计数据。根据ARI-MA模型,建立拟合值y和实际值y的相关回归模型,并计算预测未来年份的死亡数。继而将该预测值带入GRNN模型中,获得修正预测值。ARIMA运算平台采用SPSS 17.0,GRNN建模采用Matlab平台。结果经GRNN修正的ARIMA(3,2,2)模型较好拟合了1991~2005年交通伤死亡人数的时间序列,2006~2009年交通伤死亡人数的预测值符合实际值的变动趋势,预测值与实际值的相对误差率仅为5%。以之为基础预测2010~2015年的全国交通伤死亡人数为6.6~6.4万。结论 ARIMA-GRNN联合模型能较为准确的对交通伤死亡人数变化趋势做出预测。中国交通伤死亡人数已步入持续下降通道。若维持现有的道路交通安全政策水平及执法力度,交通伤致死人数将从2015年起维持于6.5万的水平。若要进一步降低致死人数水平,可能需要提前从现有法规、政策及执行基础方面进行调整。
OBJECTIVE To establish a traffic accident prediction model based on ARIMA and GRNN, and to analyze the effect of related public policies on traffic casualties Impact of a preliminary analysis. Methods The number of traffic casualties in 1991-2009 was derived from the corresponding statistics published by the Ministry of Public Security of China. According to the ARI-MA model, a correlation regression model between the fitted value y and the actual value y is established and the number of deaths predicted in the future years is calculated. Then the forecasting value is brought into the GRNN model to obtain the revised forecasting value. ARIMA computing platform using SPSS 17.0, GRNN modeling using Matlab platform. Results The ARIMA (3,2,2) model modified by GRNN fitted the time series of traffic casualties in 1991-2005 well. The predicted value of traffic casualty in 2006-2009 was in line with the trend of actual value. The forecast The relative error rate between the value and the actual value is only 5%. Based on this, it is predicted that the number of death toll from traffic accidents in 2010-2015 will be 6.6-6.4 million. Conclusion ARIMA-GRNN joint model can predict the change trend of death toll of traffic injuries more accurately. The death toll from traffic injuries in China has entered a channel of continuous decline. To maintain the existing road traffic safety policy and enforcement efforts, the number of traffic casualties will remain at 65,000 from 2015 onwards. To further reduce the number of fatalities, adjustments may need to be made in advance from the existing legislation, policies and implementation basis.