Low-rank representation based traffic data completion method

Intelligent Transportation Systems (ITS) plays a significant role in the traffic management, i.e. traffic jam prediction, route guidance. Due to the hardware failure or data transformation failure, some traffic observation data may be occasionally missed, which seriously affect intelligent transportation information service. So, the completion of traffic observation data has now become an issue that requires to be concerned and solved. By analyzing the traffic history data, we find that traffic data tend to have strong spatio-temporal correlation. Considering this feature, we propose a new low-rank representation based traffic data completion method. To further enhance the local correlation, we introduce an ordered regulation into our proposed method. We also give an efficient solution to our proposed methods. In order to verify the performance of our methods, some traffic data completion experiments are conducted on the Beijing metropolitan road speed dataset and the capital airport highway microwave dataset. Experimental results show that the proposed methods are superior to other state-of-the-art traffic data completion methods.