Traffic time series prediction based on CS and SVR

Numerous data of traffic parameters consume a large amount of storage space and take long time to be predicted. It doesn't achieve real-time performance. And compressed sensing is the algorithm that the sparse signal can be reconstructed to recover the original signal. Due to the problems, a novel CSSVR algorithm is proposed. Firstly, the theoretical analysis proves the influence to recovered error by the con-elation between measure matrix and sparse basis. And then, a general reconstruction framework is provided to reconstruct a single measurement or a multiple measurement of arbitrary sparse structure. At last, the reconstruction of the predicted sparse signal by support vector machine to get the predicted result will be explained. The simulation results show that it can realize the prediction of traffic parameter based on the sparse reconstruction efficiently and the accuracy is of high quality. So the algorithm is robust and practical.