Benchmarking Machine Learning Models for Landslide Susceptibility: A Study in the Ngororero Sector

This study evaluates the performance of six machine learning algorithms: Decision Trees (DT), Neural Networks (NN), Support Vector Machines (SVM), Random Forests (RF), Gradient Boosting Machines (GBM), and k-nearest Neighbors (k-NN) for landslide prediction in the Ngororero sector, Rwanda. Using Sentinel-2 satellite imagery, meteorological data, and topographical datasets from 2015, 2019, and 2023, the study incorporates critical features such as slope, rainfall, soil type, and vegetation cover. The findings indicate significant temporal and algorithmic variations in prediction performance. K-Nearest Neighbors and Random Forest consistently achieved high accuracies, with kNN showing a value of 84% in the 2019 dataset and more than 80% in other datasets. Random Forest demonstrated robust performance with a 78.98% accuracy in 2015. The research concludes that k-nearest Neighbors and Random Forest are optimal for predicting landslides in the Ngororero sector.