时间序列小波分解的BP神经网络AQI预测研究

AQI Prediction of BP Neural Network Based on Wavelet Decomposition of Time Series

空气质量指数预测对于人们的生活具有重要意义.针对时间序列小波分解输入BP神经网络的空气质量指数预测问题,提出了确定BP神经网络输入节点数目的方法,建立了基于时间序列小波分解输入的BP神经网络,并基于徐州地区近5 a的空气质量指数日均值数据进行了空气质量指数的建模预测仿真.结果表明,所提出的建模方法比单独使用时间序列方法和BP神经网络方法预测精度更高.该研究为BP神经网络处理时间自相关数据提供了参考.

The prediction of air quality index is of great significance to people's life.Aiming at the prediction of air quality index when time series is input into BP neural network,a method to determine the number of input nodes is proposed and the input time series are decomposed by wavelet transform.Based on the daily average data of air quality index in Xuzhou in recent five years,the prediction and simulation of air quality index are done.The simulation results show that the proposed modeling method has higher prediction accuracy than the time series method and BP neural network method alone.The research results provide a positive method reference for BP neural network to process time autocorrelation data.