Prediction and Analysis of Air Quality in Wuhan based on Improved CNN-BiLSTM Model

Accurate air quality prediction is not only helpful to public health, but also to environmental protection and urban planning. Therefore, this paper introduces the Sparrow Search Algorithm and Attention Mechanism to construct the SSA-CNN-BiLSTM-Attention model. Drawing from daily air quality data collected in Wuhan, this model initially subjects the experimental data to outlier removal and wavelet threshold denoising. Following this preprocessing, it leverages Convolutional Neural Networks (CNN) to distill meaningful features from the input data. The derived feature sequence is then channeled through a Bi-directional Long Short-Term Memory (BiLSTM) network, whose output is subsequently refined by an Attention mechanism that strategically assigns weights to salient features. Ultimately, a fully connected layer produces the prediction results, and to enhance the model's performance, the sparrow search algorithm is integrated for optimizing its parameters.The experimental results indicate that the model outperforms CNN-BiLSTM and CNN-BiLSTM-Attention models in predicting air quality in Wuhan, achieving more precise predictions and smaller errors.