Short term traffic flow prediction of urban road network driven by spatiotemporal data

To enhance the precision of predicting short-term traffic volumes within urban road systems, a novel approach propelled by spatiotemporal data is introduced in this paper. Initially, the method employs grey correlation analysis to dissect intricate nonlinear relationships inherent in short-term traffic flow data across both temporal and spatial axes of urban road networks. Subsequently, the data is segmented and undergoes greyscale transformation to yield processed inputs. These inputs are then subjected to training via an Extreme Learning Machine (ELM) neural network, culminating in the derivation of predictive outcomes through the computation of optimal output weights, thus fulfilling the objective of short-term traffic flow predicting. Empirical evidence indicates that this methodology surpasses conventional techniques in terms of both predictive accuracy and expedited predicting duration. The predictions generated align closely with the observed traffic flow patterns.