基于智能技术的交通流区间预测方法研究

发布时间：2019-06-13 17:38

【摘要】：随着城市化的不断发展,城市中交通流量的快速增长与道路基础设施增长缓慢之间的矛盾越来越凸显。尤其是特大城市的拥堵现象越来越严重,这已经严重制约我国的城市交通可持续发展战略的继续推进。智能城市交通系统(ITS)被看作是缓解这些问题的有效方法之一。因此,本文以多断面相似性为基础,讨论了交通流区间预测方法。本文研究的思路是基于已经获取的历史数据,以现有的交通流预测技术为蓝本,对交通流数据变化规律进行分析,研究其发展趋势并进行归纳,以达到对未来交通流的变化趋向进行准确预测的目的。本文主要研究工作如下:(1)针对交通流数据的内在相关性进行了分析,主要以时间序列为工具讨论并处理交通流数据的相关性及其特征。(2)针对交通流数据的断面相关性进行了分析,把两个或者多个相邻点看作一个整体,分析其临近点交通流的变化趋势以及引起变化的多种因素,提出了基于多断面相关性的预测算法,并给了相应的分析方法。(3)结合以上基于时间与空间相关性分析方法,改进了常规交通流预测算法,对其点预测结果进行置信区间计算,从而得到预测区间,提出了基于多断面相关性的交通流区间预测算法,该预测方法的模型主体是支持向量机(SVM)回归模型。(4)针对以上预测方法做了进一步的改进与增强,引入了Boosting增强算法。该算法通过使用重采样技术来进行自动的权值重置和组合,经过多次选择并训练数据,希望能够提高分类器性能。其核心思想是在训练新的分类器时,着重训练那些相对更难被正确分类的样本。
[Abstract]:With the continuous development of urbanization, the contradiction between the rapid growth of traffic flow and the slow growth of road infrastructure is becoming more and more prominent. Especially, the congestion phenomenon of mega-cities is becoming more and more serious, which has seriously restricted the sustainable development strategy of urban traffic in our country. Intelligent urban traffic system (ITS) is regarded as one of the effective methods to alleviate these problems. Therefore, based on the multi-section similarity, the traffic flow interval prediction method is discussed in this paper. The idea of this paper is based on the obtained historical data, based on the existing traffic flow prediction technology, the change law of traffic flow data is analyzed, its development trend is studied and summarized, in order to achieve the purpose of accurate prediction of the change trend of traffic flow in the future. The main research work of this paper is as follows: (1) the internal correlation of traffic flow data is analyzed, and the correlation and characteristics of traffic flow data are discussed and processed with time series as a tool. (2) the cross-section correlation of traffic flow data is analyzed, two or more adjacent points are regarded as a whole, and the changing trend of traffic flow and many factors causing the change are analyzed. A prediction algorithm based on multi-section correlation is proposed, and the corresponding analysis method is given. (3) combined with the above analysis methods based on time and space correlation, the conventional traffic flow prediction algorithm is improved, and the confidence interval of the point prediction results is calculated, thus the prediction interval is obtained, and the traffic flow interval prediction algorithm based on multi-section correlation is proposed. The main body of the prediction method is the support vector machine (SVM) regression model. (4) the above prediction method is further improved and enhanced, and the Boosting enhancement algorithm is introduced. By using resampling technology to reset and combine the weights automatically, the algorithm hopes to improve the performance of the classifiers by selecting and training the data many times. The core idea is to train the samples which are relatively difficult to be classified correctly when training the new classifiers.
【学位授予单位】：江苏大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：U491.14

【参考文献】