Optimizing Mean Fragment Size Prediction in Rock Blasting: A Synergistic Approach Combining Clustering, Hyperparameter Tuning, and Data Augmentation

• DOI: 10.3390/eng5030102
• Publication Date: 2024-08-15T09:47:11Z
• Authors: Ian Krop; Takashi Sasaoka; Hideki Shimada; Akihiro Hamanaka
• Keywords: Hyperparameter; Fragment (logic)

Accurate estimation of the mean fragment size is crucial for optimizing open-pit mining operations. This study presents an approach that combines clustering, hyperparameter optimization, and data augmentation to enhance prediction accuracy using Xtreme Gradient Boosting (XGBoost) regression model. A dataset 110 blasts was divided into 97 training testing, whereas a separate set 13 new, unseen used evaluate robustness generalization Hierarchical Agglomerative (HA) K-means clustering algorithms were used, with HA providing higher cluster quality. To address class imbalance improve model generalization, synthetic minority oversampling technique Gaussian noise (SMOGN) employed. Hyperparameter tuning conducted HyperOpt by comparing Random Search (RS) Advanced Tree-structured Parzen Estimator (ATPE). The combination ATPE SMOGN in expanded search space produced best results, achieving superior reliability. proposed HAC1-SMOGN model, which integrates tuning, augmentation, achieved squared error (MSE) 0.0002 R2 0.98 on test set. highlights synergistic benefits enhancing machine learning models tasks, particularly scenarios or limited data.