[1]潘伟靖,陈德旺.基于 GRU-SVR 的短时交通流量预测研究[J].计算机技术与发展,2019,29(10):11-14.[doi:10. 3969 / j. issn. 1673-629X. 2019. 10. 003]
　PAN Wei-jing,CHEN De-wang.Research on Short-term Traffic Flow Prediction Based on GRU-SVR[J].,2019,29(10):11-14.[doi:10. 3969 / j. issn. 1673-629X. 2019. 10. 003]
点击复制

基于 GRU-SVR 的短时交通流量预测研究()
分享到：

《计算机技术与发展》[ISSN:1006-6977/CN:61-1281/TN]

卷:
29
期数:
2019年10期
页码:
11-14
栏目:
应用开发研究
出版日期:
2019-10-10

文章信息/Info

Title:
Research on Short-term Traffic Flow Prediction Based on GRU-SVR
文章编号:
1673-629X(2019)10-0011-04
作者:
潘伟靖1; 2 ; 陈德旺1; 2
1. 福州大学 数学与计算机科学学院,福建 福州 350108; 2. 福州大学-星云股份智慧新能源研究中心,福建 福州 350108
Author(s):
PAN Wei-jing 1; 2 ; CHEN De-wang 1; 2
1. School of Mathematics and Computer Science,Fuzhou University,Fuzhou 350108,China; 2. Nebula Intelligent New Energy Research Center of Fuzhou University,Fuzhou 350108,China
关键词:
门限循环单元; 支持向量回归; 长短期记忆网络; 深度学习; 短时交通流量
Keywords:
gated recurrent unit; support vector regression; long short term memory; deep learning; short-term traffic flow
分类号:
TP31
DOI:
10. 3969 / j. issn. 1673-629X. 2019. 10. 003
摘要:
短时交通流量预测为智能交通系统(ITS)的研究方向之一。现有相关研究中所提及的基于深度学习的方法,需要较高的计算复杂度或模型的回归预测能力存在一定的不足。 因此,提出一种将门循环单元(gated recurrent unit,GRU)与支持向量回归(support vector regression,SVR)相结合的模型。 该模型一方面借助深度学习模型的强大能力进行特征提取工作,并相较长短时记忆网络(long short term memory,LSTM)降低了一定的计算量,同时,又以支持向量回归模型来增强整个模型的回归预测能力。 文中基于 Keras 所提供的 python 库,完成实验设计及开发,根据实验结果对模型进行逐步调整,选择最优模型,并在 PeMS 数据集上对模型的泛化能力以及抗噪能力进行了测试。 实验结果表明,与 SVR 以及 GRU 模型相比较,GRU-SVR 模型预测精度分别提升了 4%和 1.6%,并且该模型具有一定的泛化及抗噪能力。
Abstract:
Short-term traffic flow forecasting is one of the research directions of intelligent transportation systems (ITS). The deep learning-based approach mentioned in the existing related research requires high computational complexity or the regression prediction of the model is insufficient. Therefore,we propose a model that combines gated recurrent unit (GRU) with support vector regression (SVR). On the one hand,this model makes use of the powerful ability of deep learning model to carry out feature extraction,and reduces the computation amount compared with long short term memory (LSTM). On the other hand,support vector regression model is used to enhance the regression prediction of the whole model. The experimental design and development is completed based on the python library provided by Keras. And the optimal model is selected according to the experimental results,and the generalization and anti-noise of the model are tested on the PeMS dataset. The experiment shows that the GRU-SVR model has improved prediction accuracy by 4% and 1.6% in contrast with SVR and GRU. And the model has certain generalization and anti-noise ability.

相似文献/References:

[1]李雷 张建民.一种改善的基于支持向量机的边缘检测算子[J].计算机技术与发展,2010,(03):125.
　LI Lei,ZHANG Jian-min.An Improved Edge Detector Using the Support Vector Machines[J].,2010,(10):125.
[2]吴金花 孙德山.加权支持向量回归的权值确定方法[J].计算机技术与发展,2009,(06):135.
　WU Jin-hua,SUN De-shan.An Approach of Confirming Weight Values for Reweighted Support Vector Regression[J].,2009,(10):135.
[3]姜忠民 赵建民 朱信忠.基于最大间隔聚类的背景建模方法[J].计算机技术与发展,2009,(10):96.
　JIANG Zhong-min,ZHAO Jian-min,ZHU Xin-zhong.A Background Modeling Method of Maximum Margin Clustering[J].,2009,(10):96.
[4]张垚 袁志海 江海燕.一种面向对象软件缺陷的早期预测方法[J].计算机技术与发展,2010,(08):37.
　ZHANG Yao,YUAN Zhi-hai,JIANG Hai-yan.An Early Defect Prediction Approach for Object Oriented Software[J].,2010,(10):37.
[5]杨玫 郭天杰 陈青华.基于经验模式分解的时间序列预测方法研究[J].计算机技术与发展,2011,(07):25.
　YANG Mei,GUO Tian-jie,CHEN Qing-hua.Research on Prediction of Time Series Based on EMD[J].,2011,(10):25.
[6]宋剑,邱晓晖. 采用支持向量回归抑制噪声的经验模态分解方法[J].计算机技术与发展,2014,24(11):122.
　SONG Jian,QIU Xiao-hui. Empirical Mode Decomposition Method Using Support Vector Regression to Suppress Noise[J].,2014,24(10):122.
[7]邵曦,汪慧敏. 基于k平面分段回归的音乐情感分类[J].计算机技术与发展,2015,25(06):166.
　SHAO Xi,WANG Hui-min. Music Emotion Classification Based on k-plane Piecewise Regression[J].,2015,25(10):166.
[8]王冰玉,刘勇军.基于模糊近似支持向量回归的股价预测研究[J].计算机技术与发展,2021,31(03):14.[doi:10. 3969 / j. issn. 1673-629X. 2021. 03. 003]
　WANG Bing-yu,LIU Yong-jun.Research on Stock Price Prediction Based on Fuzzy Proximal Support Vector Regression[J].,2021,31(10):14.[doi:10. 3969 / j. issn. 1673-629X. 2021. 03. 003]
[9]何 涛,陈 剑,闻英友,等.基于堆叠模型的司法短文本多标签分类[J].计算机技术与发展,2021,31(03):27.[doi:10. 3969 / j. issn. 1673-629X. 2021. 03. 005]
　HE Tao,CHEN Jian,WEN Ying-you,et al.Multi-label Classification of Judicial Short Texts Based on Stacking Model[J].,2021,31(10):27.[doi:10. 3969 / j. issn. 1673-629X. 2021. 03. 005]
[10]孙荣荣,杨舒一,单 飞,等.基于视觉和灰度共生矩阵的医学影像质量评价[J].计算机技术与发展,2021,31(增刊):146.[doi:10. 3969 / j. issn. 1673-629X. 2021. S. 029]
　SUN Rong-rong,YANG Shu-yi,SHAN Fei,et al.Pulmonary Medical Image Quality Assessment Based on Human VisualCharacteristics and Gray Level Co-occurrence Matrix[J].,2021,31(10):146.[doi:10. 3969 / j. issn. 1673-629X. 2021. S. 029]

更新日期/Last Update: 2019-10-10