Abstract:
Rockfalls, landslides, and debris flows present significant threats to the safety of mountainous communities globally. With the rapid development of computer technology and the onset of the "big data" era, new avenues and momentum have emerged in disaster prevention and mitigation. Artificial intelligence, notably machine learning algorithms, has emerged as a hot point in this domain. Drawing upon an extensive literature review, this paper provides an overview of the application of machine learning algorithms, encompassing classical and deep learning methodologies. The current issues and future development directions are also discussed. This study highlights the critical role of classical machine learning algorithms—such as supervised, unsupervised, and reinforcement learning in assessing the susceptibility to landslides and debris flow hazards. Notably, the random forest model stands out for its high predictive accuracy and versatile modeling adaptability, making it a dependable tool for landslide susceptibility prediction. Deep learning architectures, including autoencoders, deep belief networks, convolutional neural networks, and recurrent neural networks, are instrumental in hazard identification, susceptibility assessment, and displacement prediction. Future research should prioritize enhancing data quality and quantity, optimizing model interpretability, improving model reliability and generalization, and establishing real-time monitoring and warning systems for automatic identification and rapid response to geological hazards. This study provides support and research directions for the prevention and mitigation of slope geological hazards using machine learning techniques.