Dropout-integrated attention mechanism in CNN-LSTM prediction model for vehicle-mounted air quality monitoring

The CNN-LSTM (Convolutional Neural Network-Long Short-Term Memory) model is widely used for urban air quality prediction; however, it often struggles from suboptimal accuracy due to its limited capacity to effectively capture complex spatial and temporal dependencies. To address this limitation, an attention mechanism was integrated into the CNN-LSTM model, enhancing its ability to focus on important features. However, this approach is still prone to overfitting, which compromises its generalization ability and accuracy in real-world applications. In this study, we propose the integration of dropout into the attention mechanism to enhance the CNN-LSTM model for vehicle-mounted air quality monitoring systems. The incorporation of a dropout-enhanced attention mechanism mitigates overfitting, further improving the model’s accuracy and generalization performance. Experimental results, based on real-world data collected from high-traffic urban areas, demonstrate that the Dropout Attention CNN-LSTM model outperforms both the baseline CNN-LSTM and the attention-enhanced CNN-LSTM models. Specifically, the enhanced model achieves a 6% increase in accuracy. The model also attains a precision of 95.1%, a recall of 94.2%, and an F1 score of 96.1%. These results underscore the model’s superior reliability and its suitability for real-time air quality monitoring in complex urban environments.