基于高速公路收费数据的挖掘预测分析与应用研究

发布时间：2018-10-30 20:55

【摘要】：高速公路收费数据作为高速公路联网收费管理系统的最直接产物,具有字段丰富、内容充实、数据量大、更新及时等特点。对收费数据进行深入分析及挖掘,能够得到许多隐藏在基础数据下的信息。本文基于高速公路收费流水数据的特征分析,从算法优化、模型建立、实例分析、应用对比等方面,对高速公路车辆旅行路径、断面车流量以及路段旅行时间进行预测研究,一方面改善出行者的出行选择,另一方面也能提高高速公路管理部门的管理水平。针对国内外相关预测研究中只注重单一方面,而缺乏完善且系统的综合预测挖掘研究的现状,本文主要完成了以下工作:首先,提出一种高速公路原始收费数据的预处理方法。针对收费数据中较大比例的异常数据,为最大程度减少异常数据的干扰,提出将异常数据分为冗余数据、缺失数据以及噪声数据这3类数据,分别进行处理方法说明,并通过实例对比,验证该处理方法的可行性。其次,基于马尔可夫预测法建立车辆旅行路径预测模型。分车型进行模型的建立以提高预测精度。针对预测法中状态转移概率矩阵的求解,选取统计法及线性方程组法分别进行求解并对比分析,采用实例进行验证,结果表明:统计法更适合于收费数据的预测特征。再次,在路径预测的基础上,针对高速公路路段交通状态的预测进行研究。选取路段断面车流量及车辆旅行时间作为路段交通状态的预测指标。提出一种基于收费数据的断面车流量统计法,以此为数据基础进行车流量预测。实例分析表明:基于自适应卡尔曼滤波算法的断面车流量预测,能避免卡尔曼滤波算法的缺陷,并提高预测精度。然后,论证路段旅行时间与断面车流量的相关性,提出基于密度的路段旅行时间估计方法并进行算法修正,通过MATLAB实例分析,验证该算法的可行性及准确性。最后,基于以上预测研究成果,提出高速公路收费数据的预测应用场景。分实时与非实时预测进行对比分析,并分别针对工作日、周末及节假日的不同交通特性进行交通预测,研究结果可得若干结论,用于相关领域的研究。
[Abstract]:As the most direct product of expressway network toll management system, expressway toll collection data has the characteristics of rich fields, substantial content, large amount of data, timely updating, and so on. A lot of information hidden under the basic data can be obtained by analyzing and mining the charging data in depth. In this paper, based on the characteristic analysis of expressway toll flow data, from the aspects of algorithm optimization, model establishment, case analysis and application comparison, the prediction of expressway vehicle travel path, cross-section traffic flow and section travel time are studied. On the one hand, it can improve the traveler's travel choice, on the other hand, it can also improve the management level of highway management department. In view of the fact that the related prediction research at home and abroad only pays attention to a single aspect, but lacks the perfect and systematic comprehensive prediction mining research, this paper mainly completes the following work: first, a preprocessing method of the original toll data of expressway is proposed. Aiming at the large proportion of abnormal data in charge data, in order to minimize the interference of abnormal data, the paper puts forward that the abnormal data can be divided into redundant data, missing data and noise data. The feasibility of the method is verified by an example. Secondly, the vehicle travel path prediction model is established based on Markov prediction method. The models are built to improve the prediction accuracy. In order to solve the state transition probability matrix in the forecasting method, the statistical method and the linear equation group method are selected to solve and analyze respectively. The results show that the statistical method is more suitable for the forecasting characteristics of the charging data. Thirdly, on the basis of path prediction, the traffic state prediction of expressway section is studied. The section traffic flow and vehicle travel time are selected as the predictors of road traffic state. A cross-section traffic flow statistic method based on toll data is proposed to predict the traffic flow on the basis of the data. The analysis of examples shows that the prediction of cross-section traffic flow based on adaptive Kalman filter algorithm can avoid the defects of Kalman filter algorithm and improve the prediction accuracy. Then, the correlation between section travel time and cross-section traffic flow is demonstrated, and a density-based road travel time estimation method is proposed, and the algorithm is modified. The feasibility and accuracy of the algorithm are verified by an MATLAB example. Finally, based on the above research results, the forecast application scenario of highway toll data is put forward. The traffic prediction is carried out according to the different traffic characteristics of weekdays weekends and holidays. Some conclusions can be drawn and used in the research of related fields.
【学位授予单位】：北京交通大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：U495

【参考文献】