POP-TRAFFIC: a novel fuzzy neural approach to road traffic analysis and prediction

TLDR: The obtained results highlight the capability of POPFNN-TVR in fuzzy knowledge extraction and generalization from input data as well its high degree of prediction capability as compared to traditional feedforward neural networks using backpropagation learning.

Abstract: Although much research has been done over the decades on the formulation of statistical regression models for road traffic relationships, they have been largely unsuitable due to the complexity of traffic characteristics. Traffic engineers have resorted to alternative methods such as neural networks, but despite some promising results, the difficulties in their design and implementation remain unresolved. In addition, the opaqueness of trained networks prevents understanding the underlying models. Fuzzy neural networks, which combine the complementary capabilities of both neural networks and fuzzy logic, thus constitute a more promising technique for modeling traffic flow. This paper describes the application of a specific class of fuzzy neural network known as the pseudo outer-product fuzzy neural network using the truth-value-restriction method (POPFNN-TVR) for short-term traffic flow prediction. The obtained results highlight the capability of POPFNN-TVR in fuzzy knowledge extraction and generalization from input data as well its high degree of prediction capability as compared to traditional feedforward neural networks using backpropagation learning.