Optimization of SVR functions for flyrock evaluation in mine blasting operations

This study introduces a new model to determine the critical flyrock event in mines. The flyrock was predicted and optimized using a field database including six parameters and 240 blasting events. The human learning optimization (HLO) algorithm was used in this research to optimize the support vector regression (SVR) function. Given different coefficients of kernels, optimization process minimized the likelihood of error in the models, allowing them to be detected and performed with the greatest precision. This procedure was repeated until the best model was discovered. Eventually, the radial basis function kernel was chosen for evaluating flyrock because it received the lowest computational error and the highest model accuracy. This model provided coefficient of determination (R-2) = 0.9372 and R-2 = 0.9294, respectively, as the accuracy for training and testing results. This function was considered as a relationship that the HLO algorithm could use to find the best options (i.e., optimal condition) under various conditions. The findings for 14 cases that are the essential examples in this study indicated that the optimal states are found with a great precision. The variation of the results obtained from optimization with real values is less than 5%. This demonstrates that a suitable model can be developed by employing the HLO algorithm in the development of the predictive related models to blasting and rock mechanics.