<p>Clustering is a widely used data analysis technique that groups samples based on similarity measures. Its sound statistical principles and high interpretability make it valuable across fi nance, healthcare, and earth sciences—particularly in mineral exploration. Mineral explora-tion data, often obtained through discrete sampling and grid interpolation, exhibit strong spatial dependence, with adjacent measurements generally more similar than distant ones. Conventional clustering algorithms typically ignore spatial attributes, despite the demonstrated role of spatial autocorrelation in shaping clustering outcomes. The ClustGeo algorithm, a hierarchical clustering method based on Ward’s criterion, incorporates both spatial and feature variables through a weighting parameter α to balance their contributions. In this study, we apply ClustGeo to perform spatial clustering of residual gravity anomalies by integrating spatial coordinates (X, Y) and anomaly values. This approach reduces subjectivity and improves the objectivity and interpretability of the results. Our fi ndings indicate that the Ward method produces highly cohesive clusters that eff ectively delineate regional tectonic features. However, the ClustGeo algorithm is highly sensitive to the number and type of feature variables. When only a single feature variable is used, its inherent spatial autocorrelation—combined with explicit spatial coordinates—may lead to information redundancy, diminishing the contribution of the feature variable and reducing the interpretability of clustering outcomes. Therefore, based on geochemical and residual gravity/magnetic anomaly data from the northeastern Jiaolai Basin, this study defi nes six input variables and applies the ClustGeo algorithm to predict mineral prospectivity. By analyzing the distribution of variables within each cluster, we identify prospective areas for gold and copper deposits, demonstrating the utility of spatial clustering in supporting geological research and mineral exploration.</p>