摘要
短时交通流的准确预测为智能交通系统的稳定运行提供了至关重要的技术支持。针对这一问题,提出了一种基于改进遗传模拟退火算法(IGSA)和深度回声状态网络(DESN)的短时交通流预测方法。该算法采用DESN网络对交通流量数据进行预测;并运用融合了并行搜索策略和自适应搜索策略的IGSA算法对DESN网络的超参数进行优化,得到能够使DESN网络预测性能最优的超参数值。最终利用最优超参数值构成的DESN网络对测试集数据进行预测。通过大量的对比实验表明,所提出的IGSA-DESN预测方法在短时交通流预测任务中有较好的预测性能与精准度。
The accurate prediction of short-term traffic flow provides crucial technical support for the stable operation of intelligent transportation system.A short-term traffic flow prediction method based on Improved Genetic Simulated Annealing Algorithm(IGSA)and Deep Echo State Network(DESN)was proposed to solve this problem.DESN was used to predict traffic flow data.The IGSA algorithm,which combines the parallel search strategy and the adaptive search strategy,was used to optimize the hyperparameters of DESN and obtain the hyperparameter values that can optimize the prediction performance of DESN.Finally,the DESN composed of optimal hyperparameter values was used to predict the test set data.A large number of comparison experiments show that the IGSA-DESN prediction method has better predictive and precision in the short-term traffic flow prediction task.
作者
张清勇
常万峰
李昶吾
黄荆溪
张行
ZHANG Qing-yong;CHANG Wan-feng;LI Chang-wu;HUANG Jing-xi;ZHANG Hang(School of Automation,Wuhan University of Technology,Wuhan 430070,China;School of Information Engineering,Wuhan University of Technology,Wuhan 430070,China;School of Transportation and Logistics Engineering,Wuhan University of Technology,Wuhan 430070,China)
出处
《武汉理工大学学报》
CAS
2022年第7期89-95,共7页
Journal of Wuhan University of Technology
基金
湖北省自然科学基金(2019CFB571)。
关键词
短时交通流预测
深回声状态网络
改进遗传模拟退火算法
自适应搜索策略
short-term traffic flow forecast
deep echo state network
improved genetic simulated annealing algorithm
adaptive search strategy
