Abstract: In order to predicate traffic flow and speed effectively, this paper proposes a method of predication of traffic flow based on local polynomial fitting. It obtains the flow and speed change sequence through the first order difference, uses mutual information method and false nearest neighbor method determining delay time and embedding dimension, reconstructs the state vector and curve-fit forecast function through the neighbor vectors, and uses local polynomial law. Simulation experiment has predicated actual data by new method, the results show that the estimate method is effective.

Key words: traffic flow, polynomial fitting, predication;traffic system

摘要： 为准确地预测交通流量和速度，提出一种基于局部多项式拟合的交通流预测方法。通过一阶差分得到流量和速度的变化序列，采用互信息法确定延迟时间，用虚假邻近点法确定嵌入维数，重构得到状态向量；通过邻近向量拟和预测函数，采用局部多项式预测。对实测数据进行了仿真实验，结果表明了该方法的有效性。

关键词: 交通流, 多项式拟合, 预测, 交通系统