In recent years,economic development and thecontinuous expansion of the scale of cities have dam-aged the city and its surrounding ecological environ-ment,making urban air quality problems more andmore serious.With the problem of low prediction ac-curacy of existing pollutant prediction models,a pol-lutant diffusion trend prediction method based on ahybrid network of CNN and GRU is proposed.Inthis paper,the feature set constructed by the six mainpollutant factors and climate factors in the air is usedas the input of the system model.Firstly,performingdata cleaning on abnormal data to ensure the purityof the data,and using correlation analysis to deter-mine the pollutant factor with the highest degree ofcorrelation with AQI as the prediction target.Sec-ondly,using CNN network to construct high-dimen-sional feature vectors of the time series sequence,and input the results into the GRU.Finally,trainingthe parameters of the GRU model as the output thepredicted AQI.The experimental results show thatcompared with the CNN-LSTM prediction modeland CNN prediction model,the proposed algorithmhas significant advantages in prediction accuracy.The average absolute error and root mean square er-ror are 3.365 and 2.741,respectively.The proposedmodel is O-The 50-hour long-term prediction of pol-lutant diffusion trend can also achieve better perfor-mance.
展开▼
