Research progress in landslide susceptibility assessment driven by machine learning: A bibliometric analysis

Landslide susceptibility assessment serves as a critical component in landslide hazard prevention and mitigation. With advancements in data-driven technologies, machine learning models have gained increasing prominence in this field. This study retrieved 817 peer-reviewed articles from the Web of Science Core Collection (WOSCC) and employed CiteSpace and VOSviewer for data processing and visualization to systematically investigate research progress in machine learning-driven landslide susceptibility assessment from 2009 to 2024. The results show that the number of publications in this field shows a continuous growth trend, with publications in 2024 accounting for 28% of the total number of publications. Collaborative relationships between authors were significantly strengthened after 2016, especially after 2021, showing a multi-center and high-density cooperation pattern. China dominates global scholarly output in this field with 437 publications, while forming cooperative relationships with many countries around the world. Keyword co-occurrence analysis and temporal analysis reveal a hotspot shift from traditional models to complex methods such as deep learning and ensemble learning and have begun to pay attention to model interpretability. Journals including Engineering Geology, Computers & Geosciences, Science of the Total Environment, Geomorphology, and Landslides exhibit high average co-citation frequencies, underscoring their status as platforms for high-impact research. Deep learning and ensemble learning have received increasing attention in landslide susceptibility assessment. Convolutional neural networks are increasingly seen as a promising direction for future research. Overall, this study provides an overview of the development trends, knowledge structure, and emerging research directions in machine learning-driven landslide susceptibility assessment, offering valuable insights for future research and methodological advancement.