An air quality forecast model based on the BP neural network of the samples self-organization clustering

In practice, the training samples of the neural network usually have intrinsic characteristics and regularity. The paper presents a BP neural network (BPNN) forecast model based on the samples self-organizing clustering. Using the clustering feature of the self-organizing competitive neural network(SOCNN), it improves the effect of the training sample to the performance of BPNN. The momentum - adaptive learning rate adjustment algorithm that makes the convergence speed faster with the higher error precision is used for the BPNN in this model. The experiments of the air quality forecast with this model showed that BPNN forecast model based on the samples self-organizing clustering will improve the convergence rate first and reduce the possibility of falling into the local minimum also and improve the forecast accuracy.