Soft-Computing Techniques Applied to Short-Term Traffic Flow Forecasting

Multi Linear Perceptron (MLP) neural networks, Radial Basis Function (RBF) neural networks, and Fuzzy Logic (FL) were used as soft-computing (or artificial intelligent) modelling techniques to come to short-term forecasts of traffic flow. The field data used for modelling was collected through single loop induction detectors on freeways 405 and 22 in Orange County, California. The data consisted of 30-s time bins of traffic flow [vehicles/30 s], occupancy [seconds/30 s], and reliability indicators (value 0-7) that were transformed into 5-min time bins. The paper concludes that the MLP and the RBF(2) - with 1 output - method outperform the FL and the RBF(1) - with 2 outputs method.