基于时空模型的交通流故障数据修正方法
Repair method of traffic flow malfunction data based on temporal-spatial model

为了提高交通流数据的准确性,从时间相关性、空间相关性和历史相关性三方面分析了交通流大数据的特点,建立了基础交通流时空模型。为保证数据处理的精度和速度,进行了时空模型的简化和标定。将时空模型简化,抽象为双层规划模型,上层模型通过控制时空相关参数的数量实现运算速度的优化,下层模型通过控制误差实现计算精度的优化。应用数据驱动法进行双层规划模型的求解,完成时空模型的标定。在时空模型的基础上,提出了交通流故障数据修正方法。以北京市某路段为例,对交通流故障数据修正方法进行有效性和可行性验证。验证结果表明:基于历史趋势、空间相关与时间序列的交通流故障数据修正方法的精度分别为79.65%、85.16%、89.84%,基于时空模型的交通流故障数据修正方法的精度为90.91%,具有较高的精度,而且可准确描述交通流大数据的特点。
In order to improve the accuracy of traffic flow data, the characteristics of temporal correlation, spatial correlation and historical correlation of traffic flow big data were considered, and a basic traffic flow temporal-spatial model was built. To ensure the accuracy and speed of data processing, the simplification and calibration of temporal-spatial model were realized. The temporal-spatial model was simplified and abstracted into a bi-level programming model. The operation speed was optimized in the upper level model by controlling the number of temporal-spatial correlation coefficients, and the calculation accuracy was optimized in the lower level model by controlling the error. Based on the data-driven method, the bi-level programming model was solved, and the calibration of temporal-spatial model was completed. Based on the proposed temporal-spatial model, a repair method of traffic flow malfunction data was presented. Taking a road section in Beijing City as example, the validity and feasibility of proposed repair method were verified. Verification result indicates that the precisions of repair methods of traffic flow malfunction data based on historical trend, spatial correlation and time series are 79.65%, 85.16%, 89.84% respectively, however, the precision of proposed method based on the temporal-spatial model is 90.91%, that is relatively higher, the characteristics of traffic flow big data can be accurately described by the proposed repair method. 2 tabs, 19 figs, 25 refs