An urban and rural spatial governance effectiveness evaluation system based on AI and multimodal data

Effective spatial governance in urban and rural areas is crucial for sustainable development, optimized resource allocation, and improved public service delivery. Conventional evaluation methods, relying on limited datasets and manual assessments, often fail to capture the complexity of modern spatial systems. This research proposes an Adaptive Star Fish Optimized Multilayer Perceptron Network framework integrating multimodal data to evaluate spatial governance, identify underperforming regions, and provide actionable insights for policy and planning. An Urban and Rural Spatial Governance dataset of 1,000 records across China was compiled, measuring governance effectiveness with a composite score based on expert-weighted infrastructure, services, environment, satisfaction, and economic metrics. Data sources included census records, Internet of Things sensor data, and social media feedback. Preprocessing involved tokenization, z-score normalization, and temporal-spatial alignment. Features were extracted via Bag-of-Words for textual data and Wavelet Transform for numerical signals and fused at the feature level to form a comprehensive representation of governance conditions. Performance was evaluated to enhance prediction accuracy, identify critical factors, and rank intervention priorities. Implemented in Python, the framework achieved 87.1% prediction accuracy. SHapley Additive exPlanations interpreted key feature contributions. The system manages heterogeneous data, captures spatiotemporal variations, and provides a scalable, interpretable tool for targeted urban-rural interventions.