基于数据流挖掘的交通流状态辨识方法研究

An Effective RTRC-TFD Method for Detecting Traffic Flow Using Data Stream Mining

由于交通流的复杂性,使得基于交通流参数的交通流状态经典辨识算法的阈值设定十分困难,直接影响辨识效果和效率。文章根据交通流的特性,提出了一种自适应动态学习的分类算法RTRC-TFD,可将其应用于不同背景下交通流的数据流实时分类。实验结果表明:在考虑概念漂移和背景重现的条件下,RTRC-TFD相对于经典检测算法(增量式贝叶斯分类算法)具有更高的分类精度和更快的收敛性。

Aim. The introduction of the full paper points out what we believe to be the four characteristics of traffie flow that cause difficulties in determining the thresholds of massive traffic flow; in its last effective RTRC-T17D （recognizing and treating recurring context of traffic flow detection） method, which is ex- plained in section 1. The seven subsections of section 1 are： （1） traffic flow data stream model, （2） classification model, （3） time for context extension, （4） determination of context recurrence, whose detection algorithm framework is given in Fig. 1, （5） pattern of context change, （6） time for context extension, （7） algorithm for recogniz- ing and treating recurring context, whose procedural steps are given in Fig. 2： To verify the effectiveness of our method, Section 2 uses the measurement data in a report by U.S. Department of Transportationtgl to simulate our algorithm; the simulation results, presented in Figs. 3 through 9, and their analysis show preliminarily that when concept drift and recurring context are taken into consideration, our algorithm has better classification precision and convergence speed, thus being more effective for classifying real-time traffic flow data stream from various contexts than the conventional incremental Bayes classification algorithm.