摘要
利用主成分分析方法对路网中各路口的交通流量进行了相关性分析 ,构建了相关路口集 ,提出了根据相关路口集的交通流量预测本路口流量的思想 ,给出了用于预测的神经网络模型、具体算法和评价标准。在比较精确的训练样本基础上对网络进行了训练 ,测试了训练后的网络性能。实验表明 。
In this paper, an effective method of dynamic traffic-flow prediction based on dependency analysis of urban intersections traffic-flow, is proposed to researc h on relations of the varying traffic-flow. The correlation of intersections tr affic-flow is studied by PCA (Principal Component Analysis); accordingly the co rrelative intersection set is built from what traffic-flow of objective interse ction can be predicted. The model, algorithm and evaluation standard of the ANN (Artificial Neural Network) are also given for prediction. After training the ne twork with quite accurate samples, the performance of ANN is tested. Experimenta l results have shown that the method of traffic-prediction based on dependency a nalysis of urban intersections traffic-flow has satisfying accuracy and better robustness.
出处
《交通与计算机》
2005年第1期31-34,共4页
Computer and Communications
