<p>Precision management of soil lead (Pb) in peri-urban areas requires accurately identifying pollution sources and their spatially varying influences. However, traditional land-use data are too coarse to pinpoint high-risk industries, and widely used ensemble learning models typically fail to explicitly reveal spatially varying driver effects, which is crucial for actionable environmental zoning. To address these gaps, this study developed a framework integrating Points of Interest (POI) and PM2.5 data for fine-scale characterization of industrial sources, and applied the Geographical Pattern Interaction (GPI) model for spatial partitioning and dominant factor identification. This framework was applied to Huangpi District, Wuhan city, China, a typical peri-urban area undergoing rapid industrialization. The GPI model demonstrated superior performance, significantly outperforming random forest and extreme gradient boosting in predictive accuracy (predicted R2 = 0.688) and effectively mitigating overfitting. Crucially, the model generated an optimal spatial partition, delineating six homogeneous subregions characterized by distinct Pb contamination level and dominant controlling factors. This partition revealed a clear spatial gradient: from northern uplands dominated by PM2.5 to southern industrial cores controlled by heavy industry emissions derived from POI kernel density, with transitional zones showing mixed influences. Furthermore, the GPI model quantified strong, spatially varying local interactions, particularly between industrial emissions and soil properties. Leveraging these mechanistic insights, we precisely delineated priority control zones and proposed targeted, threshold-based mitigation strategies. This work demonstrates that integrating POI data with the GPI model provides a powerful and interpretable framework for dominant factor identification and zoning management of soil heavy metals in complex peri-urban landscapes.</p>