Unequal Interval Dynamic Traffic Flow Prediction with Singular Point Detection

Analysis of traffic flow signals plays an important role in prediction and management. As intrinsic property, the singular point a signal labels new nonsteady status. Therefore, detecting is effective approach to determine moment prediction. In this paper, improved wavelet transform proposed detect points real-time signals. The number detected output via heuristic selection multiple scales. Then, weighted similarity measurement historical utilized predict next point. position decides duration adaptively. predicted are applied dynamically update unequal interval flow. Furthermore, Vasicek model used by minimizing sum relative mean standard error (RMSE) between increment sampled increments previous intervals. A decomposition method solve matrix problem. Based on scenario imported from real-world map, simulation results show that algorithm outperforms existing approaches with high accuracy much lower computing cost.