With rising levels of air-pollution, air-quality forecasting has become integral to the dissemination of human health advisories and the preparation of mitigation strategies. Traditional air quality models, such as the Community Multi-scale Air Quality (CMAQ) model, have unsatisfactory accuracy. Accordingly, a correction model, which combines convolutional neural network (CNN) and long-short term memory neural network (LSTM) and optimized by harris hawks optimization algorithm (HHO) was established to enhance the accuracy of CMAQ model's prediction results for six air pollutants (SO_2, NO_2, PM_10, PM_2.5, O_3 and CO). The accuracy of HHO-CNN-LSTM was evaluated using root mean square error (RMSE), mean absolute error (MAE), and the index of agreement (IOA). The results demonstrated a significant improvement in the accuracy of prediction for the six pollutants using the correction model. RMSE decreased by 73.11% to 91.31%, MAE decreased by 67.19% to 89.25%, and IOA increased by 35.34% to 108.29%. To address the propensity of the HHO algorithm to converge on local optima, leading to poor CO correction performance, this study proposed a method for the HHO algorithm with a Gaussian random walk strategy to improve the CO concentration correction performance.