In this study， we propose an innovative method for developing a regional seismic economic loss prediction model by integrating regional building characteristic parameters with ground motion intensity measures， enabling rapid generation of large-scale seismic loss data to provide reliable data support for regional seismic risk assessment. The research workflow comprises three core modules： foundational data construction， data augmentation， and model development.（1） For foundational data construction， a seismic damage simulation framework is established based on a simplified multidegree-of-freedom structural model considering soil-structure cluster interaction， achieving highprecision calculation of building component damage and loss by following FEMA P-58.（2） For data augmentation， an improved data augmentation algorithm is proposed to address the bottleneck of insufficient sample data in traditional methods by generating high-fidelity simulated loss data through feature space reconstruction and distribution optimization techniques. （3） In model development， a regional loss prediction model based on Gaussian process regression is established， which effectively characterizes spatial correlation mechanisms of building cluster losses by incorporating feature variables， such as building functional type and spatial distribution. Empirical studies conducted on a university campus demonstrate that， compared with traditional data augmentation methods， the improved algorithm considerably enhances the fidelity of simulated data， and that neglecting interbuilding loss correlations may result in the underestimation of total regional loss predictions. Through the integration of physical-and data-driven approaches， this model effectively resolves prediction biases due to sample insufficiency and neglected correlations in traditional data augmentation methods while maintaining computational efficiency， providing a new technical pathway for urban seismic resilience assessment.