Abstract: The peak vibration velocity of open-pit mine throw blasting is a critical parameter for assessing both the safety of blasting operations and their environmental impact. To enhance the accuracy of peak vibration velocity prediction,this study employed Spearman and Kendall correlation coefficients for statistical analysis. Furthermore, by integrating Random Forest (RF) and Extreme Gradient Boosting (XGBoost) algorithms,the most influential features affecting peak vibration velocity were identified. Based on these findings,blast center distance,height difference,bench height,total charge,and average unit consumption were selected as the input variables for the prediction model. The weighted average method was applied to combine the outputs of the XGBoost and improved particle swarm optimization hybrid kernel extreme learning machine(IPSO-HKELM) models,forming an integrated prediction framework. Results indicate that by fusing the predictions of these two models using the weighted average approach and adjusting the distribution of sample weights,the overall performance of the model was significantly enhanced. The determination coefficient (R 2 ),mean absolute error (MAE),root mean square error (RMSE) and mean absolute percentage error (MAPE) of the weighted average ensemble model were 0. 977,0. 591,0. 921 and 17. 198%,respectively. Compared with the traditional method,the weighted average method integration model shows a significant improvement in the evaluation index,especially in the MAE and RMSE,which shows its advantages in practical application. 

Key words: throwing blasting,XGBoost,HKELM,weighted average method