Abstract: Blasting fragmentation is an important factor affecting the production efficiency and cost control of open-pit
mines. Accurate prediction of fragmentation is of great significance for optimizing the design of blasting parameters. To this
end,an XGBoost hybrid integration model based on Alpha Evolutionary Algorithm (AEA) optimization is proposed. The model
uses AEA to optimize the hyperparameters of the XGBoost model through 97 sets of blasting sample data,and combines crossvalidation
to evaluate its performance. The results show that the AEA-XGBoost model performs well in prediction accuracy,stability
and generalization ability. The highest determination coefficient of the model is 0. 921 4,the variance interpretation rate is
92. 75%,and the root mean square error is 0. 045,which is significantly better than the traditional artificial neural network,Knearest
neighbor algorithm,gradient boosting decision tree,LightGBM and AEA optimized random forest model,showing stronger
robustness and anti-overfitting ability. In order to further improve the interpretability of the model,the Shapley additive interpretation
algorithm is used for analysis,and it is revealed that the elastic modulus,the ratio of blockage length to resistance
line and the original rock fragmentation are the key characteristics affecting the prediction of blasting fragmentation. Finally,
based on 8 actual blasting engineering cases,the application effect of AEA-XGBoost model is verified. The predicted value is
highly consistent with the measured value,which proves the feasibility and effectiveness of the model in complex geological conditions
and actual blasting environment.

Key words: hybrid ensemble model,AEA optimization algorithm,fragmentation size prediction,Shapley model explanation