Optimized Air Quality Prediction Model Based on Neural Network

Air pollution is still the biggest factor jeopardizing environmental public health and safety. In fact, the model based on artificial intelligence research has been widely used in actual air quality prediction. However, for the special field of meteorology, the closeness of the prediction result from a single value (such as minimizing the root mean square error) does not necessarily indicate a good prediction effect, and the analysis of experimental results should be multifaceted. When building a model, it often encounters the occurrence of spatiotemporal instability, delay effects, overfitting, etc., which makes the prediction accuracy need to be improved. To overcome these bottlenecks. Based on the current research status of deep neural networks, this paper proposes a highly accurate air quality prediction model based on data mining. In terms of efficiency, combined with the complexity of the parameters in the meteorological field, a method based on the weighted limit learning machine (W-ELM) parameter learning is proposed. The weights of the input layer and the random assignment of the offset can be calculated. Learning speed has been greatly improved. In terms of accuracy, an algorithm combining sparse autoencoder (SAE) and swish activation function is proposed, which can greatly improve the prediction accuracy. Through experiments, it can be proved that the model has high fitting degree and low distortion rate, which can provide better reference value for air prediction research.