<![CDATA[Landslides are among the most frequent natural disasters in Indonesia, posing significant threats to human safety and infrastructure, particularly in regions characterized by steep topography and high rainfall intensity. This study aims to spatially analyze landslide susceptibility using machine learning algorithms and to identify the most influential conditioning factors contributing to landslide occurrences. The research was conducted in Kabupaten Luwu Utara, located in Sulawesi Selatan, utilizing recorded landslide events from 2015–2025. The analysis incorporated biophysical and geospatial variables, including rainfall, humidity, and the Normalized Difference Vegetation Index (NDVI). The study employed a quantitative explanatory approach with predictive modeling based on Random Forest (RF), Support Vector Machine (SVM), and Artificial Neural Network (ANN) algorithms. The dataset was constructed as a balanced binary classification of landslide and non-landslide events. Model performance was evaluated using multiple metrics, including accuracy, precision, recall, F1-score, ROC-AUC, and precision–recall curves to ensure comprehensive assessment. The findings indicate that the Random Forest model achieved the highest overall accuracy, while the Artificial Neural Network demonstrated superior capability in detecting landslide occurrences within the testing dataset. However, the relatively modest AUC and Average Precision values suggest limitations in the models’ discriminative performance. Variable importance analysis revealed that rainfall is the most dominant conditioning factor, followed by humidity and NDVI. Overall, the machine learning approach successfully represents landslide susceptibility spatially. Nevertheless, further model optimization and integration of additional relevant variables are required to enhance predictive reliability and to support robust disaster mitigation and early warning systems.]]>
