This paper investigates landslide susceptibility in Constantine, Algeria, using advanced machine learning techniques integrated with Geographic Information Systems (GIS) and remote sensing data. The study employs two robust gradient boosting algorithms, XGBoost and LightGBM, to develop predictive models. These models are fine-tuned using Bayesian optimization, which enhances their accuracy. The model performance is validated through cross-validation and interpreted using SHAP values, providing a detailed analysis of the factors influencing landslide risk. The study shows high predictive accuracy, with ROC AUC scores consistently exceeding 0.83, highlighting the effectiveness of the applied methods. Key factors influencing landslide susceptibility, such as elevation, slope, and proximity to roads, are identified, offering critical insights for risk assessment and management in landslide-prone areas. This research underscores the potential of integrating machine learning and GIS for effective natural hazard prediction and urban planning strategies.