基于灰色关联分析和SOM-RBF神经网络的高速公路交通量预测方法研究  

Prediction Method of Expressway Traffic Volume Based on Grey Correlation Analysis and SOM-BRF Neural Network

作　　者：毛锦伟 宋伟炜 周昱辰 MAO Jin-wei;SONG Wei-wei;ZHOU Yu-chen(Three Gorges University,Yichang,Hubei 443002,China;Hunan University of Science and Technology,Xiangtan,Hunan 411100,China;Rongcheng Yongkang Ocean Engineering Consulting Co.Ltd.,Weihai,Shandong 264200,China)

机构地区：[1]三峡大学,湖北宜昌443002 [2]湖南科技大学,湖南湘潭411100 [3]荣成市永康海洋工程咨询有限公司,山东威海264200

出　　处：《广东交通职业技术学院学报》2020年第3期10-15,共6页Journal of Guangdong Communication Polytechnic

摘　　要：本文选取两段高速公路交通量相关数据作为样本,基于聚类算法改进RBF神经网络对交通量进行预测,考虑影响因素的复杂多样,其中对云南元磨高速普洱段数据建立灰色关联分析,得到选取的特征影响因素的关联度大小,其中普洱市总人口数关联度最小为0.5532。随后引入自组织特征映射神经网络(SOM)构建聚类模型。采用先聚类分析、再分别预测的思路,解决了由于RBF神经网络对于少量样本和训练样本点分散所引起的预测精度降低的问题,改进的神经网络泛化能力有所提高,结果表明:SOM-RBF组合算法对元磨高速交通量进行预测,其相对误差维持在6%以下,平均相对误差为3.81%,预测效果较BP神经网络和RBF神经网络有较大的提升。通过两段高速公路的实例分析,验证了SOM-RBF组合算法有良好的预测效果和适用性,可有效的用于交通量预测,具有较高的实用价值。In this paper,the traffic volume data of the two expressways are selected as samples,and the traffic volume is predicted based on the improved RBF neural network and clustering algorithm.Considering the complexity and diversity of the influencing factors.the gray correlation analysis is established for the data of Yunnan Yuanmo high speed Pu’er section,and the correlation degree of the selected characteristic influencing factors is obtained.The minimum correlation degree of the total population of Pu’er City is 0.5532.Then self-organizing feature mapping neural network(SOM)is introduced to construct the clustering model.With the idea of clustering analysis first and then forecasting separately,this paper solves the problem that the prediction accuracy of RBF neural network is reduced due to the dispersion of a small number of samples and training samples,and the generalization ability of the improved neural network is improved.The results show that the relative error of SOM-RBF combined algorithm in the prediction of traffic volume of Yuanmo high speed is kept below 6%,and the average relative error is 3.81%.The prediction effect is better than that of BP neural network and RBF neural network.Through the example analysis of two freeways,it is proved that SOM-RBF combination algorithm has good prediction effect and applicability,can be effectively used for traffic volume prediction,and has high practical value.

关 键 词：自组织特征映射神经网络 径向基函数 灰色关联分析 交通量 

分 类 号：U491.14[交通运输工程—交通运输规划与管理]