基于频繁模式的短时交通流实时预测研究

发布时间：2018-03-03 13:09

本文选题：智能交通　切入点：短时交通流　出处：《南京邮电大学》2017年硕士论文　论文类型：学位论文

【摘要】：近年来,智能交通系统蓬勃发展,已经成为了人们日常出行的可靠保障。短时、实时的交通流预测是实现交通管理和车辆诱导的重要基础,也是智能交通系统领域的一个重要研究内容。大数据时代的发展也带来了交通数据的爆炸式增长,在海量的交通流大数据的背景下,如何做到实时准确的预测成为了一个新的问题。本文对高速路采集的大规模数据集下短时交通流的实时预测进行了深入的研究、分析和实验。主要的研究工作有以下几点:(1)为了提高挖掘和预测模型的准确性,需要对采集的交通数据流信息进行预处理,本文针对这一特定的数据进行分析,进行了缺失填补、错误修正和符号离散化等ETL(Extract-Transform-Load)处理。(2)在当前大数据环境的背景下,提出了一种能够实时挖掘交通流数据频繁闭合模式的算法—TP-Moment(Moment of Topology)。算法基于Topology的并行计算模型对传统的Moment算法进行了改进。实验证明,在大规模数据集的环境下,算法的时间和空间等性能得到很大提升,能够快速准确地挖掘出所有有效模式,很好的满足交通流预测的准确性和实时性要求。(3)在TP-Moment算法的基础上,提出了基于历史频繁模式的短时交通流预测模型。挖掘历史交通流数据存在的频繁模式规则,与获得的实时交通流信息进行最近邻规则的匹配,从而预测未来时刻的交通流状况。实验证明,该预测模型具有较高的预测精度,有效可行。
[Abstract]:In recent years, Intelligent Transportation system (its) has developed rapidly and has become a reliable guarantee for people's daily travel. In the short term, real-time traffic flow prediction is an important basis for traffic management and vehicle guidance. It is also an important research content in the field of intelligent transportation system. The development of big data era has also brought about the explosive growth of traffic data. How to achieve real-time and accurate prediction has become a new problem. In this paper, the real-time prediction of short-time traffic flow under the large-scale data set collected by highway is deeply studied. In order to improve the accuracy of mining and prediction models, we need to preprocess the collected traffic data flow information. The missing fill, error correction and symbol discretization are processed by ETLL Extract-Transform-Load.2) in the context of big data's current environment, This paper presents an algorithm for real-time mining frequent closure patterns of traffic flow data. Based on the parallel computing model of Topology, the algorithm improves the traditional Moment algorithm. The performance of the algorithm, such as time and space, has been greatly improved, which can quickly and accurately mine all the effective patterns, and meet the accuracy and real-time requirements of traffic flow prediction. (3) on the basis of TP-Moment algorithm, This paper presents a short-term traffic flow prediction model based on historical frequent pattern, mining frequent pattern rules of historical traffic flow data, matching the obtained real-time traffic flow information with nearest neighbor rules. The experimental results show that the model has high prediction accuracy and is effective and feasible.
【学位授予单位】：南京邮电大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：U491.14;TP311.13

【相似文献】