Document Type : Original Research Paper
Authors
Department of Science Education, Farhangian University, P.O. Box 14665-889, Tehran, Iran
Abstract
Anomaly detection is the process of recognizing patterns in data that differ from the typical behavior. In geochemistry, this involves identifying hidden patterns and unusual components within the context of exploratory target identification. This issue is particularly significant when limited information is available about the area of interest. Therefore, employing methods that can aid in the exploration process under such conditions and with limited data is highly valuable. In this study, the Deep-Embedded Self-Organizing Map (DE-SOM), an unsupervised deep learning approach, was used to detect geochemical anomalies. The research focused on identifying multivariate geochemical anomalies in the Moalleman region. After detecting the region's geochemical anomalies, the effectiveness of the algorithm was assessed alongside two other types of SOM algorithms. For this purpose, the prediction area plot was utilized, with the intersection points for DE-SOM, Batch SOM, and SOM were determined to be 0.75, 0.67, and 0.65, respectively. The multivariate geochemical anomaly in the Moalleman area shows a good correlation with known mineral occurrences and the andesite and dacite units. Based on this, it can be stated that the DE-SOM method is a useful tool for identifying anomalies and patterns associated with mineralization.
Keywords
Main Subjects
)