Comparison of unsupervised clustering methods in distinguishing mineralized catchment basins

The paper focuses on the use of unsupervised clustering methods to identify and characterize mineralized catchment basins in the Jebal Barez region in southeast Iran, known for its copper deposits. The study utilized a dataset of 658 stream sediment samples that were collected and analyzed for 20 elements. Different clustering approaches – K-means, fuzzy C-means, hierarchical, and spectral clustering – were employed to evaluate, compare and contrast their efficiency in distinguishing and characterizing mineralized catchment basins. To evaluate the performance of the clustering techniques, different clustering validation criteria such as the Caliński–Harabasz Index, Silhouette Index, and within-cluster sum of squares were utilized to determine the optimal number of clusters and to evaluate the outcomes of different clustering approaches. The study found that K-means clustering was the superior method for accurately distinguishing different clusters. Of the 16 known mineral occurrences in the study area, 11 were recognized by two K-means clusters. It was also found that K-means clusters 2 and 3 describe the most promising zones for copper mineralization. Based on these findings, the study demonstrated the potential of unsupervised clustering methods for effective exploration targeting in the region.

Journal of Geochemical Exploration

Journal of Geochemical Exploration

292

0375-6742

N/A

N/A

10