TRAINING NEURAL NETWORKS TO DETECT FREEWAY INCIDENTS BY USING PARTICLE SWARM OPTIMIZATION
ridership - commuting
Particle swarm optimization, Neural networks, Multilayer feed-forward neural networks, Motorways, Incident detection, Freeways, Controlled access highways, Backpropagation, Artificial neural networks, ANNs (Artificial neural networks), Algorithms
Among the many models and techniques developed to automatically detect freeway incidents in recent years, the multilayer feed-forward neural network (MLF) is one of the most promising models in terms of high detection rate, low false alarm rate, and faster mean time to detection. The use of particle swarm optimization (PSO) algorithms to train MLFs to detect freeway incidents is investigated in an attempt to further improve detection performance. Several MLFs have been trained by different variations of the PSO methods using real incident data from Interstate 880 in California. The best MLFs were compared with one trained by the conventional backpropagation (BP) algorithm. The evaluation discussed, based on I-880 data, shows that the MLFs trained by the PSO algorithms have the same or higher detection rates, similar or lower false alarm rates, and faster mean time to detection than the MLF trained by the BP algorithm. This research has shown that PSO has the potential to improve the good incident detection performance of the MLF.
Cheu, R, Srinivasan, D, Loo, W. (2004). TRAINING NEURAL NETWORKS TO DETECT FREEWAY INCIDENTS BY USING PARTICLE SWARM OPTIMIZATION. Transportation Research Record, Vol. 1867, p. 11-18.