Optimization of support vector machine parameters in modeling of Iju deposit mineralization and alteration zones using particle swarm optimization algorithm and grid search method

The support vector classifier (SVC) is one of the most powerful machine learning algorithms. This algorithm has been accepted as an effective method in three-dimensional geological modeling. Although the model selection has a great impact on the performance of SVC algorithm, most of mining studies have neglected it and used the grid search method. Therefore, in this study, a new approach is proposed for improving the selection of SVC models. This approach uses particle swarm optimization (PSO) to determine the important parameters of SCV such as penalty and kernel parameters. The proposed approach was applied in the modeling process of the Iju porphyry copper deposit to delineate alteration and mineralization zones. The optimal penalty and kernel parameters were found to be 2 7.2 and 2 −4.75 for alteration zone, and 2 2.72 and 2 −6.23 for mineralization zone, respectively. With 97.4% and 97.01% rates of accuracy for mineralization and alteration zones, the PSO results showed reasonable performance in classification. The proposed approach had better accuracy than grid search method. Therefore, because of its better performance, the geological models were developed using the PSO method to be used as a basis for future resource evaluation. • The particle swarm optimization (PSO) algorithm and grid search method were compared in parameter optimization. • Implicit geological modeling by a support vector classifier whose parameters were optimized using PSO. • Implicit geological model of alteration and mineralization domains are separately prepared.