摘要
为了研究可适用于高原山区高速公路短时交通流的预测方法,以及预测方法思路对绩效的影响,提出基于广义回归神经网络(General Regression Neural Network,GRNN),构建K均值聚类算法(K-means clustering algorithm,K-means)与GRNN混合预测方法思路,即通过K-means和绩效指标判断GRNN模型参数最佳值,进而建立最佳预测模型。与传统上通过经验或一定指标判断模型参数值的思路相比,采用K-means和GRNN混合预测思路得出的模型参数值更佳,且模型RMSE、MAE最高可分别改善45.92%、45.05%,则构建的混合预测方法思路是科学有效的,可为高原山区交通流预测方法优化提供借鉴。
In order to study the short-term traffic flow prediction method which can be applied to the expressway in plateau and mountains areas,and the impact of prediction method idea on performance,a mixed prediction method idea of k-means clustering algorithm(K-means) and GRNN is proposed based on the general regression neural network(GRNN),that is,the best value of GRNN model parameters is judged by K-means and performance indicator,and then the best prediction model is established.Compared with the traditional idea of judging the model parameter value through experience or certain indicators,the model parameter value obtained by using the K-means and GRNN mixed prediction idea is better,and the RMSE and MAE of the model can be improved by 45.92% and 45.05% respectively,so the mixed prediction method idea constructed is scientific and effective,and can provide reference for the optimization of traffic flow prediction method in plateau and mountainous areas.
作者
林美
梁艳洁
陆彬
LIN Mei;LIAGN Yanjie;LU Bin(Institute of Transportation Development Strategy&Planning of Sichuan Province,Chengdu Sichuan 610041,China)
出处
《交通节能与环保》
2024年第2期67-73,共7页
Transport Energy Conservation & Environmental Protection
基金
四川省省级科研院所基本科研业务费支持项目(2021JBKY05)。
关键词
运输规划与管理
短时交通流预测
GRNN
K-MEANS
高原山区高速公路
transportation planning and management
short-term traffic flow prediction
GRNN
K-means
expressway in Plateau Mountainous areas
