基于异常值识别卡尔曼滤波器的短期交通流预测

Outlier-identified Kalman filter for short-term traffic flow forecasting

针对智慧交通的需求提出了一种新颖有效的短时交通流预测方法,通过异常值识别扩展了卡尔曼滤波,使其能对噪声进行识别和过滤——异常值识别卡尔曼滤波器。利用卡尔曼滤波能有效地过滤导致系统不确定性的交通流波动,但这可能会使指示交通流突变的细微线索丢失,为了提升预测精度,应用离散小波变换对原始信号进行识别处理,在去掉异常值的同时保留原有对预测有效的信号源信息,此外还使用了历史参考值对预测值进行修正。在四个基准数据集上的大量实验表明,与常用及最新的预测模型相比,其结果MAPE平均降低了2.919%,RMSE平均降低了79.582。

This paper presented a novel and effective technique for short-term traffic flow forecasting.The main contribution was an extension of Kalman filter,such that it became to be able to identify the outlier and then filter out it,which was named as outlier-identified Kalman filter.It used the classic Kalman filter to filter out the fluctuations of the traffic flow which led the system uncertainty,and the subtle clues indicating the sudden change of the traffic flow might lose in this operation.To achieve better forecasting accuracy,it used discrete wavelet transform to process the original signal and preserved the useful signals while de-noising.In addition,it applied historical reference values to correct predicted values.Extensive experiments on four benchmark datasets demonstrate an average improvement of 2.919%in MAPE and 79.58 in RMSE.