Accurate and reliable air quality predictions are critical to the ecological environment and public health. For the traditional model fails to make full use of the high and low frequency information obtained after wavelet decomposition, which easily leads to poor prediction performance of the model. This paper proposes a hybrid prediction model based on data decomposition, choosing wavelet decomposition (WD) to generate high-frequency detail sequences WD(D) and low-frequency approximate sequences WD(A), using sliding window high-frequency detail sequences WD(D) for reconstruction processing, and long short-term memory (LSTM) neural network and autoregressive moving average (ARMA) model for WD(D) and WD(A) sequences for prediction. The final prediction results of air quality can be obtained by accumulating the predicted values of each sub-sequence, which reduces the root mean square error (RMSE) by 52%, mean absolute error (MAE) by 47%, and increases the goodness of fit (R²) by 18% compared with the single prediction model. Compared with the mixed model, reduced the RMSE by 3%, reduced the MAE by 3%, and increased the R² by 0.5%. The experimental verification found that the proposed prediction model solves the problem of lagging prediction results of single prediction model, which is a feasible air quality prediction method.

The accurate prediction of PM_2.5 concentrations is important for environmental protection. The accuracy of the commonly used prediction methods is not high; so, this paper proposes a PM_2.5 concentration prediction method based on a hybrid learning model. The Top-of-Atmosphere Reflectance (TOAR), PM_2.5 data decomposed by wavelets, and meteorological data were used as input features to build an integrated prediction model using random forest and LightGBM, which was applied to PM_2.5 concentration prediction in the Beijing–Tianjin–Hebei region. The practical application showed that the proposed method using TOAR, incorporating wavelet decomposition with meteorological element data, had an improvement of 0.06 in the R² of the model accuracy and a reduction of 2.93 and 1.14 in the root mean square error (RMSE) and mean absolute error (MAE), respectively, over the model using Aerosol Optical Depth (AOD). Our model had a prediction accuracy of R² of 0.91, which was better than the other models. We used this model to estimate and analyze the variation in PM_2.5 concentrations in the Beijing–Tianjin–Hebei region, and the results were the same as the actual PM_2.5 concentration distribution trend. Obviously, the proposed model has a high prediction accuracy and can avoid the errors caused by the limitations of the AOD inversion method.