Detection of multivariate geochemical anomalies using machine learning (ML) algorithms in Dehaq Pb-Zn mineralization, Sanandaj-Sirjan zone, Isfahan, Iran

The identification of geochemical anomalies is of great importance in mineral exploration as it plays a crucial role in the comprehensive exploration process. Big data analytics that utilize entire datasets and variables provide a powerful approach to identify multivariate geochemical anomalies through machine learning algorithms. These algorithms are effective in capturing complex relationships between geochemical features and mineralization. However, challenges such as data redundancy, computational complexity and high dimensionality in anomaly detection need to be overcome. Regression-based machine learning algorithms such as Isolation Forest (IF) and One-Class Support Vector Machine (OCSVM) as well as hybrid models combining Robust Principal Component Analysis (RPCA) with IF and OCSVM were used to identify Pb and Zn geochemical anomalies. Hierarchical Clustering (HC) was used for initial grouping of the data and the effectiveness of the anomaly detection methods was evaluated using Receiver Operating Characteristic (ROC) curves and prediction area (P-A) plots. A strong spatial correlation between detected anomalies and known mineral occurrences was observed, especially with the hybrid RPCA + IF model, which outperformed the other methods in terms of accuracy and reliability in anomaly detection. The results demonstrate the effectiveness of combining robust data preprocessing techniques with machine learning for accurate geochemical anomaly detection, making the RPCA + IF approach a promising tool for mineral exploration. In this study, advanced methods for geochemical anomaly detection are presented that utilize the strengths of machine learning and data preprocessing to handle the complexity of high-dimensional geochemical data. This provides a more accurate and efficient framework for the identification of Pb and Zn mineralization in the Dehaq region and increases the potential for broader applications in mineral exploration.