Method for defective traffic flow data mending based on SARBF neural networks

Defective data to collect urban traffic flow message are always occurred due to the sensor failure. To mend the defective data, a new approach named SACRBF neural network fitting is presented. It combines analysis based on spatial autocorrelation and RBF neural network fitting method. The complete data is determined to mend the defective data according to the spatial autocorrelation of traffic grid. Not only the mending precision is improved and also the limitation of regression analysis is avoided by using RBF neural network. Finally, the experiment to mend the defective traffic flow data in Hangzhou is shown that the method is practicable.