Gold anomaly classification using supervised machine learning algorithms including a novel fuzzy kernel-based extreme learning machine

• A new type-2 fuzzy kernel-based extreme learning machine (FKELM) method is introduced. The model utilizes type-2 fuzzy sets in its design process. • The FKELM performs well compared to standard and ensemble machine learning classifiers, including MLP, SVM, random forest, and gradient boosting. • The performance of all classifiers, including FKELM, is assessed with 10-fold cross-validation and ROC-AUC analysis. It is demonstrated that the FKELM is a robust classifier that distinguishes between fertile and unfertile samples and is efficient for gold anomaly detection. • Ten geochemical elements (As, Fe, S, Te, Sb, Cu, Co, Ni, Zn, W) are analyzed to detect gold anomalies. Arsenic (As), iron (Fe), and sulfur (S) are found to be the most important elements that are associated with gold ores. • The study advances gold exploration by enabling efficient classification of gold anomalies through ML models, including the FKELM. Determining effective indicator elements is crucial for identifying areas with potential gold deposits or pinpointing gold-rich zones, as well as for detecting alterations in rocks that are found in gold ores. The prediction of the gold deposits relies on these indicator elements associated with gold mineralization. Several algorithms, such as Multi Layer Perceptron (MLP), Support Vector Machine (SVM), Random Forest, Decision Tree, Gradient Boosting, K-Nearest Neighbor, and Ensemble models, are analyzed to classify geochemical anomalies in gold data and examine gold structures. To enhance classification accuracy and pinpoint elements that help in identifying gold-rich zones, an innovative Fuzzy Kernel-based Extreme Learning Machine (FKELM) model is developed. The FKELM incorporates the principles of type-2 fuzzy sets along with the Hemachor t -norm in the framework. The comparative analysis shows that FKELM achieves superior performance (AUC = 0.94 and accuracy 85.70%) to all the methods. The ROC-AUC results highlight that FKELM models exhibit superior accuracy and robustness compared to individual classifiers. These results highlight the FKELM model’s effectiveness in accurately classifying geochemical anomalies and its potential for identifying areas with potential gold deposits.