Abstract:

With the development of traffic informatization, increasing amount of traffic data can be collected. How to make most of the traffic data to forecast traffic flow is a crucial work of the intelligent transportation systems (ITS). To resolve this problem, this paper proposes a sampling method based on the combination of stratified sampling method and kmeans clustering. The support vector machine (SVM) based on sequence optimization method is used to forecast traffic flow parameters. The analysis results show that the sampling quality based on the proposed sampling method is reformed. The forecasting precision based on the SVM also gets improved. It proves that the method is efficient to solve largescale traffic forecasting problems.

Key words: intelligent transportation system, traffic congestion identification, sampling, k-means clustering

摘要：

随着交通信息化的快速发展，可供分析的交通流数据量越来越大，如何利用大规模交通流数据进行交通预测分析是智能交通的重要研究内容.为解决大规模交通流数据预测问题，本文提出了一种基于分层抽样与k均值聚类相结合的抽样方法，并与基于序贯最小优化方法的支持向量机结合，进行大规模交通流预测.实例分析结果表明，本文提出的聚类方法比现有抽样方法的抽样质量有所提高，基于序贯最小优化方法的支持向量机可有效提高交通流预测的精度.因此，本文提出的方法对于大规模交通流预测是有效的.

关键词: 智能交通, 拥挤判别, 抽样, k均值聚类