基于加权组合模型的短时交通流预测研究

为了改善城市路网中短时交通流预测效果,提高预测精度,设计了一种基于改进的K近邻非参数回归和小波神经网络加权组合的短时交通流预测方法.针对K近邻非参数回归预测方法搜索量大、相似性差等问题,采用基于交叉口相关系数加权的欧氏距离选择K近邻值.小波变换与神经网络有机结合形成的前馈型网络,对非平稳的输入信号能够呈现出良好的时频特性和变焦能力,对短时交通流预测效果有着明显的提升.通过算例分析,说明所设计的预测方法能够获得比较精确的短时交通流预测结果.

Research on Short-Term Traffic Flow Prediction Based on Weighted Combinational Models

In order to improve the effect and the prediction accuracy of short-term traffic flow prediction in urban road network,a short-term traffic flow prediction method based on a weighted combination method of improved K neighbor nonparametric regression and wavelet neural network(WNN) is designed.In order to solve the problems about K neighbor nonparametric regression prediction method,such as large amount of search and poor similarity,the Euclidean distance based on intersected weighted correlation coefficient was used to choose the K neighbor values.The feedforward network formed by the combination of wavelet transform and neural network can present a good time-frequency characteristic and zoom capability for non-stationary input signals,and has a obvious promotion for short-term traffic flow prediction.The example shows that the designed prediction method can obtain accurate results of short-term traffic flow prediction.