Back Propagation Neural Network in the Water Quality Evaluation of Qingdao Dagu River

Compared with other methods, NN (neural network) model in water quality evaluation had better performance in the application to water quality assessment. In this paper, In order to improve the NN model's performance, arithmetic, determination of hidden layer nodes amount and the training samples were optimized. Gradient descending arithmetic added by momentum and self-adaptive learning rate was chosen. The amount of nodes in network's hidden layer was optimized by pilot calculation arithmetic based on empirical equation. Training samples was extended by random differential in critical value space of grades to improve model's robustness and veracity of distinguishing. When it was used to evaluate the water quality of Qingdao Dagu river, the improved ANN (artificial neural network) model displayed a good performance.