ABSTRACT
Air quality is heavily dependent on local weather condition[1]. Under dierent weather conditions, the concentration of pollutants on the ground caused by the same pollution emissions can vary several times or even hundreds of times and artificial neural network (ANN) is the eective tool to describe and characterize the nonlinear phenomenon, but on the condition that the amounts of data are small, ANN lacks the precision of prediction. In addition, its training is slow and is easy to fall into local optimum. Taking advantages of the gray system under a smaller amount of data, we establish a gray neural network and applied genetic algorithm optimization to the forecast of the air pollution index.
