<![CDATA[The objectives of this study are: 1) Spatial modeling of landslide prediction, 2) Analysis of factors that contribute most to landslide occurrence, and 3) Landslide prediction model as a basis for landslide disaster mitigation planning. This study uses a multivariate statistical approach with random forest machine learning, and model validation is performed by calculating the Area Under Curve (AUC) value using the R language. The variables analyzed in this study include landslide locations, slope direction, slope curvature, slope inclination, elevation, rainfall, land cover, geology, soil type, landform, distance from roads, distance from rivers, and vegetation index. The results of the study found 57 landslide locations spread across the study area. The resulting machine learning random forest produced a landslide hazard prediction map with an AUC value of 0.9062, classified into five hazard level categories based on probability values: very low with an area of 5,240.98 hectares (27.64%), low with an area of 4,468.25 hectares (23.56%), moderate with an area of 4,336.05 hectares (22.86%), high with an area of 3,048.19 hectares (16.07%), and very high with an area of 1,870.42 hectares (9.86%). The largest contributing factors to landslides were slope gradient, rainfall, and distance from roads. Mitigation strategies based on the primary contributing factors to landslides include the construction of retaining walls, soil retention structures, drainage improvements, revegetation of slopes with strong root systems, and regular monitoring.]]>
