Mining activities often cause mining-induced ground deformation, including subsidence and landslides, and related geo-environmental impacts, posing significant risks to infrastructure and safety. This study conducts a systematic assessment to identify, categorize, and evaluate AI-based methods (machine learning, deep learning, and hybrid models) for predicting and monitoring mining-induced ground deformation. The literature search was performed across major scientific databases, using predefined keywords and selection criteria, resulting in a final dataset of relevant peer-reviewed studies. The reviewed works were classified into three methodological groups: traditional machine learning, deep learning-based approaches, and hybrid methods. The results show that ML still dominates in terms of accuracy and explainability, DL has the potential to handle big data and real-time prediction, but requires large datasets and computational resources, and Hybrid combines the advantages of ML and DL to increase the efficiency of prediction and risk assessment. Besides, the study highlights the strengths and weaknesses of each approach and suggests future research directions. By systematically classifying and comparing AI-based prediction and monitoring approaches in terms of data requirements, accuracy, interpretability, and real-time capability, the presented results provide practical guidance for mine managers and engineers in selecting appropriate models for deformation risk assessment and in developing effective early warning systems for safer and more sustainable mining operations.
NGUYEN et al. (Tue,) studied this question.