Geochemical Anomaly Detection and Pattern Recognition: A Combined Study of the Apriori Algorithm, Principal Component Analysis, and Spectral Clustering

This study demonstrates the effectiveness of combining Principal Component Analysis (PCA) and the Apriori algorithm for feature selection, alongside Spectral clustering, to detect geochemical anomalies in Mississippi Valley-Type (MVT) Pb-Zn deposits in western Iran. First, PCA and Apriori enabled the identification of both syngenetic and epigenetic components, which helped in recognizing elements associated with mineralization. These elements were then modeled using Spectral clustering to detect geochemical anomalies. Unlike traditional methods like k-means, Spectral clustering does not require spherical clusters and is adept at identifying clusters of arbitrary shapes. This made it particularly suitable for analyzing the irregular shapes of geochemical anomalies in the study area. By incorporating Spectral clustering, the method effectively separated geochemical groups, revealing the underlying structure of the data. This was crucial for identifying anomalous geochemical zones and delineating areas with a high potential for Pb-Zn mineralization. The performance of the Spectral clustering algorithm was thoroughly evaluated using the Silhouette Score, the Davies–Bouldin Index, and Dunn Index. Subsampling was employed to assess the algorithm’s stability, providing a comprehensive evaluation of its effectiveness in identifying geochemical anomalies and mapping mineralization potential.