Three-dimensional (3D) geological models are essential for geological analysis and mineral resource estimation. Although conventional on-site survey methods, such as boreholes, provide local engineering geological information for 3D geological modeling, accurately predicting strata in areas with sparse borehole data remains a challenge. This study proposes a 3D geological modeling method using the Transformer model under the conditions of sparse borehole data. First, a K-dimensional tree was used to identify boreholes adjacent to the target point, and a borehole context sequence was constructed using stratigraphic information from neighboring boreholes. Subsequently, the relationship between the target point and its adjacent borehole sequence was calculated using the multi-head attention mechanism of the Transformer model. Finally, trained Transformer encoders were used to predict the stratigraphic category of the target point, and the normalized information entropy was used to quantify uncertainty during the modeling process. Experimental results showed that the accuracy of the method was 0.86, outperforming the accuracy and uncertainty of a recurrent neural network. The root mean square error is smaller than the inverse distance weight and Kriging. Compared to other methods, the proposed method can more accurately describe the geometric shape and distribution of geological bodies and reveal the sedimentary laws of the study area.

Geochemical measurements of stream sediments are practical for small-scale mineral exploration. However, traditional grid interpolation methods cause element concentrations to diffuse and smooth out anomalies, particularly in complex terrains, making it challenging to reflect the actual distribution of elements accurately. We applied the Dynamic Enhanced Weighted Drainage Catchment Basin (DE-WDCB) method to enhance the retention and identification of local anomalies by limiting the scope of analysis to specific drainage units. This method reduces interference from varying background values across different watersheds, effectively enhancing geochemical element anomalies and aligning better with geomorphic conditions. The DE-WDCB method was tested in the Duobaoshan–Heihe area, a significant copper polymetallic mineral district in northeastern China. Compared with traditional grid interpolation methods, the DE-WDCB method retained and strengthened low and weak abnormal information of favorable mineralization elements, particularly in the Luotuowaizi area. The method demonstrated a higher spatial coverage rate with mineral points and a more vital ore-indicating ability. Specifically, the DE-WDCB method identified anomalies with a mean accuracy of 63.57% (p < 0.05, 95% CI: 47.64%–79.50%), compared to 50.53% for traditional methods. In conclusion, in regions with a complex topography and watershed differences, the DE-WDCB method effectively reduces local geochemical background interference, accurately identifies low and weak geochemical anomalies, and better reflects the actual distribution of elements. This makes it a significantly advantageous method for geochemical anomaly extraction, delineating higher-confidence exploration targets in the Sandaowan–Luotuowaizi area in the east and the triangular area between Duobaoshan, Yubaoshan, Sankuanggou, and the midstream highlands of the Guanbird River in the west.