Improving pillar stability assessment in underground mining using a novel SSA-GBC framework

Pillar stability is a critical factor in underground mining engineering. It plays a vital role in ensuring both structural integrity and operational safety. This study proposes a novel approach to enhance prediction of pillar stability using the Gradient Boosting Classifier (GBC), optimized through the Sparrow Search Algorithm (SSA). The assessment of pillar stability was conducted using GBC, which focused on five essential factors: the width of the pillar, its height, the ratio of width to height, uniaxial compressive strength (UCS), and the stress experienced by the pillar. These findings highlight the value of combining ensemble learning techniques with nature-inspired optimization methods to develop effective and intelligent decision-support systems. The standard GBC model achieved an accuracy of 85%. To improve the performance more, SSA was applied to optimize the hyperparameters of the GBC model, which improved the accuracy to a significant level, reaching 94%, and enhanced recall, precision, and F1-score values across both stable and unstable classes. A confusion matrix and receiver operating characteristic curves (ROC) were applied to analyze and further confirm the model performance and accuracy, particularly for unstable pillars, proving the model’s practical reliability.