基于时空相关性的交通流故障数据修复方法

为及时对高速公路交通流故障数据进行有效修复，综合考虑交通流数据的时空特性，提出基于3D形函数的时空插值修复方法.以时间间隔、距离和时滞参数作为相关数据的提取依据，以高速公路实际数据对所提出方法进行验证；将实验结果与采用时间序列法、空间插值法、基于灰色残差GM模型以及基于统计相关分析的方法得到的结果进行对比.结果表明，该方法的修复结果优于时间序列法和空间插值法，并且修复误差低于其他方法.其中，与基于灰色残差GM模型和基于统计相关分析的方法相比，该方法的修复结果的均绝对误差分别降低了21.33%和43.54%，均方根误差分别降低了12.87%和35.08%.该方法的修复结果的平均绝对值误差率比基于统计相关分析的方法降低了40%.这表明研究中所提方法的修复精度更高，是一种有效的数据修复方法.
Abstract: Considering the spatial-temporal characteristics of the traffic flow data, a spatial-temporal interpolation repair method based on 3D shape function was proposed to effectively repair the fault data of freeway traffic flow in time. The time interval, distance and time delay parameters were chosen as the extracted evidences of the relevant data, and the proposed method was validated through the actual data of freeway; while, the time series method, the spatial interpolation method, the method based on residual error GM model and the method based on statistical correlation analysis were selected as comparative approaches. Results show that the repair results of the proposed method are better than the results by time series method and spatial interpolation method; in addition, the repair error is lower than other methods. Compared with the method based on residual error GM model and the method based on statistical correlation analysis, the absolute error of the proposed method are reduced by 21.33% and 43.54%, respectively; the root-mean-square error are reduced by 12.87% and 35.08% respectively. The average absolute error rate of the proposed method are reduced by 40% compared with the method based on statistical correlation analysis, which illustrates that the repair precision of the proposed approach is more accurate and it is a kind of effective fault data repair approach.