An Improved Back Propagation Neural Network Forecasting Model Using Variation Fireworks Algorithm for Short-time Traffic Flow

Short-term traffic flow has complicated non-linear characteristic, and it has a similar seasonality. This paper presents an improved Back Propagation neural network for Short-time traffic flow prediction using Variation Fireworks Algorithm. In the new model, we use self-adaption explosion amplitude to resolve the contradiction between local search and global search while maintaining the good properties of better fireworks. Furthermore, this model, using self-variation mutation operator, can improve the features of better fireworks. For increasing the diversity, we introduce random mapping out-of-boundary sparks instead of modulo operator. It is applied to real-world data collected from the Second Ring Road of Beijing, China and is compared with three traffic prediction models. The consequences show that the new model has higher accuracy and can be served as forecasting of short-term traffic flow. And the model shows better result than the Back Propagation (BP) neural network model, Particle Swarm optimization Back Propagation (PSO-BP) neural network and Fireworks Algorithm Back Propagation (FWA-BP) neural network model.