<p>The identification of landslide risk aids in accurately mapping susceptibility, which is essential for disaster relief operations and sustainable land use planning. Due to inaccessible landslide distributions, typical detection techniques are mostly ineffective because of their considerable workload, subjectivity, and inaccuracy. This research proposes an explainable machine learning strategy for landslide risk identification that addresses these constraints. To improve the identification of different possible landslide types, the method is built on a multistage fusion architecture that integrates knowledge about probable landslide-causing factors. First, an information gain filter-based feature selection method is used to select the most relevant features and train several linear and ensemble classification models using the filtered features. After that, it makes a landslide prediction and explains, respectively. When compared with other machine learning models, this ensemble model provides better performance in landslide prediction tasks. Also, the proposed model uses some effective visualization techniques for accurate explainability tasks.</p>