Abstract: With the rapid development of big data and artificial intelligence technologies, geoscience research is increasingly shifting toward a data-driven paradigm. The integration of multi-source geological information with intelligent algorithms for mineral resource prediction has become a major focus in mineral exploration. Taking the Taxi−Baitazi Area in Inner Mongolia Autonomous Region as the study area, this study integrated 1∶50 000 geological, geophysical data, and geochemical data and constructed a sample dataset using a sliding window technique. A two-dimensional convolutional neural network （CNN2D） was then employed to establish an intelligent tin ore prospecting prediction model. By comparing the feature differences between known mineralized windows and background areas, the model identified favorable metallogenic zones across the entire study area. After parameter optimization, the optimal model was obtained and applied to prospecting prediction. The results show that the favorable prospecting areas delineated by the model account for approximately 11.42 % of the total study area. After excluding known ore occurrences, three additional prospecting prediction areas are identified. Comprehensive analysis indicates that these predicted areas show a good spatial correspondence with the major ore-controlling factors in the Taxi−Baitazi Area, indicating that the prediction results of the established model are reliable.