Prediction of ozone concentration using back propagation neural network with a novel hybrid training algorithm

This paper proposed a new hybrid forecasting model for the prediction of ozone concentrations in semi-arid area. It is based on chaotic, particle swarm optimization algorithm (CPSO) and back propagation (BP) neural network, called CPSO-BP neural network. The results show that the proposed hybrid model is superior to both the BP neural network and the regression model being tested. The hybrid model achieves 18.7% in root mean square error reduction compared to BP model, and 8.1% reduction compared to stepwise regression model. It could be a promising model on forecasting ozone concentration in semi-arid area.