Prediction of air-overpressure caused by mine blasting using a new hybrid PSO-SVR model

The aim of the present study is to predict air-overpressure (AOp) resulting from blasting operations in the Shur river dam, Iran. AOp is considered as one of the most detrimental side effects induced by blasting. Therefore, accurate prediction of AOp is essential in order to minimize/reduce the environmental effects of blasting. This paper proposes a new hybrid model of particle swarm optimization (PSO) and support vector regression (SVR) for AOp prediction. To construct the PSO-SVR model, the linear (L), quadratic (Q) and radial basis (RBF) kernel functions were applied. Here, these combinations are abbreviated using PSO-SVR-L, PSO-SVR-Q and PSO-SVR-RBF. In order to check the accuracy of the proposed PSO-SVR models, multiple linear regression (MLR) was also utilized and developed. A database consisting of 83 datasets was applied to develop the predictive models. The performance of the all predictive models were evaluated by comparing performance indices, i.e. coefficient correlation (CC) and root mean square error (RMSE). As a result, PSO can be used as a reliable algorithm to train the SVR model. Moreover, it was found that the PSO-SVR-RBF model receives better results in comparison with other developed hybrid models in the field of AOp prediction.