摘要
为避免因主观因素确定的平滑系数对预测结果产生影响,首先基于MATLAB对传统三次指数平滑法进行改进,实现了平滑系数α的动态测定;其次,为克服由于数据的随机波动性对预测结果产生影响,再次采用马尔科夫模型对上述预测结果进行修正,进一步提高了预测结果的精度;最后,以1996-2012年全国交通事故死亡总人数为基础数据进行实证分析,其中动态测定的平滑系数α为0.596;同时以-9,-3.5,2,7.5,13为划分阈值建立4个状态区间,对2013-2015年的预测结果进行马尔科夫修正,研究发现：未修正前预测结果的平均绝对误差为2.1%,修正后平均绝对误差为：1.21%,精度提高了0.89%,由此表明修正结果能更好的对实际情况作出解释并具有工程实用性.
In order to avoid the influence of the smoothing coefficient determined by the subjective factors on the prediction results, this paper firstly improves the traditional cubic exponential smoothing method based on MATLAB, and realizes the dynamic measurement for the smoothing coefficient. Secondly, so as to overcome the random fluctuation＇s effect on the prediction results, the Markov Model is used to correct the prediction results above and further the accuracy of the prediction results is promoted. Finally, the empirical results are based on the number of nationwide traffic accidents from 1996 to 2012, the coefficient of dynamic measurement is 0.596. At the same time, four state intervals are established with the threshold of-9, -3.5, 2, 7.5, and 13 to amend the prediction results previous from 2013 to 2015 with Markov Model. The results show that mean absolute error is 2.1% before correction, after amendment, being 1.21%, and the accuracy is improved by 0.89%, which means the correction result can explain the actual situation and the methods have practicality.
作者
李守金
崔子梓
赵静
LI Shou-jin;CUI Zi-zi;ZHAO Jing(School of General Education,Weifang University of Science and Technology,Weifang 262700,China;School of Mechanical Engineering,North University of China,Taiyuan 030051,China;School of Economics and Management,Chang＇an University,Xi＇an 710064,China)
出处
《数学的实践与认识》
北大核心
2018年第12期169-177,共9页
Mathematics in Practice and Theory
基金
山东省软科学研究计划项目(2017RKA07001)
关键词
三次指数平滑
马尔科夫模型
交通事故预测
预测方法
cubic exponential smoothing
Markov model
traffic accident prediction
prediction method
