考虑交通吸纳点的非参数回归组合型短时交通流预测方法

The Combined Forecasting Approach Based on Non-Parametric Regression for Short-term Traffic Flow of Roads with Parking Spaces

在城市路网中,交通吸纳点是非常重要的影响交通流量的因素。由于交通吸纳点增加了流量预测的复杂度,一般基于模型的智能算法也不能很好地解决有吸纳点的交通流预测问题。提出一种基于非参数回归的组合方法PCA-FC-NPR,即在对原始数据进行主成分分析PCA(P rincipal Com ponen t A nalysis)和模糊聚类FC(Fuzzy C lustering)的基础上,采用以数据驱动的k-近邻非参数回归方法NPR(N on-Param etric R egression)加以解决。针对交通吸纳点与检测器的位置关系采取不同预测算法。短时交通流预测示例结果证明该方法的有效性。

Parking space is a important factor in city traffic flow. Due to its much complication, normal intellectualized algorithms are not competent. A combined method of PCA-FC-NPR is presented. It adopts PCA（Principal Component Analysis）and FC（Fuzzy Clustering）based on original data. Finally the data-driven method of NPR（Non-Parametric Regression） is used. The different forecasting approaches are adopted for parking space and detectors having different locations. A demonstration is given to show its availability.