ABSTRACT
Landslides pose a significant risk to mountainous communities, particularly in the Himalayas. This study evaluates machine learning algorithms for predicting landslide suscepti-bility in Dharche Rural Municipality, a region near the epicenter of the 2015 Gorkha Earthquake in Nepal. The research employs Random Forest (RF) and Logistic Regression (LR) models to analyze 15 Landslide Conditioning Factors (LCFs), including topography, geology, and land use. A dataset comprising 571 landslide events was divided into training (80%) and validation (20%) sets. The LR model achieved an accuracy of 85% and area under the curve (AUC) value of 0.90, while RF attained 86% with AUC value up to 0.92. These findings can support policymakers in disaster risk reduction, resource allocation, and infrastructure planning. By enhancing landslide prediction, this research contributes to mitigating disaster impacts and fostering resilience in cen-tral Himalayan region of Nepal.
