<![CDATA[Floods and landslides are recurrent hydrometeorological hazards that cause significant environmental damage and socioeconomic losses in many regions of Indonesia, including North Sumatra. Complex topography, high rainfall intensity, land-use changes, and rapid urban development have increased the exposure and vulnerability of several districts to these disasters. This study aims to classify the severity of flood- and landslide-affected areas in North Sumatra using an integrated Geographic Information System (GIS) and Random Forest (RF) approach. The research was conducted using the CRISP-DM framework, which includes data collection, data preprocessing, feature weighting using the Analytical Hierarchy Process (AHP), model development with the RF algorithm, and spatial validation using historical disaster records. Five main conditioning factors were used as model inputs: rainfall, slope, land cover, soil type, and elevation. Hazard severity was classified into three categories: low, moderate, and severe. The results indicate that the RF model achieved strong predictive performance, with high precision, recall, F1-score, and an excellent ROC-AUC value, demonstrating the reliability of the proposed approach. Spatial analysis shows that Mandailing Natal, South Tapanuli, and Humbang Hasundutan are the most severely affected districts, mainly due to high rainfall, steep slopes, and land degradation. This study concludes that the GIS–RF framework provides an effective decision-support tool for regional disaster risk management and can support evidence-based planning for flood and landslide mitigation in North Sumatra.]]>
