Air quality prediction based on a DBO-LSTM model

To deal with the problem of insufficient precision in existing air quality forecasting models, a novel predictive framework is introduced. The Dung Beetle Optimizer (DBO) approach is utilized in this structure to optimize the hyperparameters of Long Short-Term Memory (LSTM) neural networks. By doing so, it surmounts the constraints typically linked with conventional manual parameter adjustment, which often relies on empirical experience. First, the DBO algorithm is employed to enhance the LSTM network’s hyperparameters. Second, we construct a DBO-LSTM model using air quality data from Kunming, inputting the best parameter set to enhance the precision of predictions. Ultimately, we evaluate our model's performance in comparison with the autoregressive integrated moving average LSTM model. The simulation outcomes reveal that the LSTM model enhanced by the DBO algorithm surpasses the baseline models when evaluated using root mean square error and mean absolute error metrics, measuring 3.4904 and 4.4902 μg/m3, respectively, with a goodness of fit reaching 90.17%. Compared with the three algorithms, whale optimization algorithm, gray wolf optimizer, and particle swarm optimization, the DBO algorithm has more obvious effects when optimizing the LSTM network for the task of air quality prediction.