[Objective] The role of urban expansion in driving landscape fragmentation through complex human-land interactions represents a critical research topic in sustainable urban planning. [Methods] To reveal its spatial patterns and driving mechanisms, this study takes the Beijing-Tianjin-Hebei region as a case study. Data from three time points (2000, 2010, and 2022) and nine natural and anthropogenic factors were employed. An integrated analysis was conducted using spatial autocorrelation, the XGBoost-SHAP model, and the PLS-SEM model. [Results] The results indicate that: ①Landscape fragmentation in the Beijing-Tianjin-Hebei region generally showed an intensifying trend, with significant spatial clustering effects. High-fragmentation hotspots were mainly concentrated in the northwestern Yanshan Mountains and along the Taihang Mountains; ②The XGBoost-SHAP model revealed that the importance ranking of driving factors for landscape fragmentation is as follows: Land Cover Change Intensity > Slope > Soil Organic Carbon Content > Human Activity Intensity > DEM > Annual Precipitation > Mean Annual Temperature > Impervious Surface Expansion Intensity > Human Footprint. Among these, Land Cover Change Intensity, Slope, Annual Precipitation, and Mean Annual Temperature primarily exerted positive driving effects on landscape fragmentation, whereas Soil Organic Carbon Content and Human Activity Intensity showed negative effects; ③PLS-SEM path analysis further clarified the action pathways of each factor and their interactions. Land Cover Change Intensity had a direct positive impact on landscape fragmentation, while Slope, Soil Organic Carbon Content, DEM, Annual Precipitation, Mean Annual Temperature, and Impervious Surface Expansion Intensity mainly exerted indirect effects by influencing Human Activity Intensity, Land Cover Change Intensity, and Human Footprint. [Conclusion] Through multi-model integration, this study clarifies the characteristics and driving mechanisms of landscape fragmentation in the Beijing-Tianjin-Hebei region, supplementing theoretical research on fragmentation in urban agglomerations. The results provide a new paradigm integrating "machine learning attribution" and "structural equation validation" for analyzing the complex causes of landscape fragmentation, offering an analytical pathway and methodological reference for studying regional landscape patterns.