期刊论文

Short-term traffic flow prediction based on the feature information extraction applying ASNBD-CMFE

中文

北京交通大学学报

1673-0291

2021

45

2

28-35

10.11860/j.issn.1673-0291.20200082

National Key R& D Plan (2018YFF0212902); National Natural Science Foundation of China(51805161); Research Key Project of Hunan Provincial Department of Education (16A180).

本校为其他机构

针对短时交通流数据的非线性和随机性特点,为提高其预测精度,提出了一种基于自适应最稀疏窄带分解（Adaptive Sparsest Narrow-band Decomposition, ASNBD）和复合多尺度模糊熵（Composite Multiscale Fuzzy Entropy, CMFE）的短时交通流数据特征信息提取方法.首先利用ASNBD将短时交通流数据分解成若干个内禀窄带分量,分别求出每个分量的CMFE.根据CMFE反映的不同分量的非线性程度选取有效分量,从而提取数据的非线性特征.然后根据非线性分析的结果分别对每个分量建立支持向量机网络模型,针对每个分量的自身特点选择不同的模型训练参数,以提高单个模型预测精度.最后将各个预测值进行累加并得出预测...

The short-term traffic flow data are nonlinear and random. In order to improve the prediction accuracy, a method for the feature extraction of short-term traffic flow data based on the correlation analysis of Adaptive Sparsest Narrow-Band Decomposition (ASNBD) and Composite Multiscale Fuzzy Entropy (CMFE) is proposed. First, the short-term traffic flow data are decomposed into several intrinsic mode narrow-band components using ASNBD. The CMFE of each component is calculated. The effective components are selected according to the nonlinear degr...