Abstract: Conventional tin exploration and prospectivity assessment rely heavily on extensive field geological surveys and are often constrained by low efficiency, limited effectiveness, and inadequate reuse of existing data. The rapid development of artificial intelligence provides new technical approaches for integrating existing geological and mineral exploration data and conducting efficient, data-driven mineral prospectivity assessment. This study focuses on the Baiyinkundi area in Inner Mongolia, China, and integrates multi-source heterogeneous geoscience datasets, including 1:50,000 geological, geochemical, and geophysical data. A mineral prospectivity model based on a deep residual network (ResNet) was developed and trained to identify areas favorable for tin mineralization. The results show that the training process converged stably and that the resulting ResNet model achieved an accuracy of 99.64 % on the test set. Application of the trained model delineated three prospective target areas in the Baiyinkundi area. These results provide technical support for further tin exploration in the study area and offer a methodological reference for strategic mineral exploration.