A CNN-PGL integrated learning model for urban vitality prediction based on multisource data

Urban vitality prediction is crucial for sustainable urban planning and smart city development. However, existing models often struggle with data heterogeneity, nonlinear relationships, and insufficient interpretability. This study proposes a novel CNN-PGL (Convolutional Neural Network-Particle Swarm Optimized Gradient Boosting Decision Tree-LightGBM) integrated learning framework to address these challenges. The model combines the nonlinear fitting capability of the Gradient Boosting Decision Tree (GBDT), the computational efficiency of LightGBM, and the deep feature extraction ability of the Convolutional Neural Network (CNN), and it is enhanced by Particle Swarm Optimization (PSO) for hyperparameter tuning. Experimental results demonstrate that the CNN-PGL model achieves superior prediction accuracy compared to standalone models like GBDT, LightGBM, and PSO-GBDT. Additionally, a feature importance analysis is conducted to identify key factors influencing urban vitality changes, enabling the efficient allocation of urban resources. The proposed model serves as a robust tool for urban vitality monitoring and decision-making, providing a scientific basis for urban planning.