This study integrates machine learning ensembles for landslide susceptibility mapping (LSM) in northern Pakistan. The research area, located in the Central Karakoram National Park, is prone to landslides due to rugged topography, frequent seismic events, and seasonal rainfall. The study combines baseline models (logistic regression, K-nearest neighbors, support vector machine) with ensemble algorithms (random forest, LightGBM, XGBoost, AdaBoost, Dagging, and Cascade Generalization) to identify the most significant factors influencing landslides. Using 228 landslide inventory maps, the study employed a random forest classifier and correlation-based feature selection (CFS) to determine the twelve most important parameters, including slope, elevation, aspect, geological features, and proximity to faults, roads, and streams. Slope was identified as the primary factor, followed by aspect and rainfall. The models were validated with metrics such as AUC, ACC, and Kappa, with XGBoost showing the highest performance (AUC 0.907, ACC 0.927, K 0.620) and logistic regression (LR) as the most accurate baseline model (AUC 0.784, ACC 0.912, K 0.394). The results highlight the effectiveness of LSM in guiding risk-mitigation strategies in geohazard-prone regions. The study demonstrates that integrating machine learning models enhances the accuracy and reliability of LSM, providing valuable insights for disaster management and policy-making. The final LSM was stratified into five susceptibility levels: very low, low, moderate, high, and very high, based on the natural breaks technique. The findings underscore the importance of slope, aspect, and rainfall in landslide prediction and the potential of ensemble methods in improving predictive accuracy. The study contributes to the understanding of landslide susceptibility factors and the application of machine learning in geohazard risk assessment.This study integrates machine learning ensembles for landslide susceptibility mapping (LSM) in northern Pakistan. The research area, located in the Central Karakoram National Park, is prone to landslides due to rugged topography, frequent seismic events, and seasonal rainfall. The study combines baseline models (logistic regression, K-nearest neighbors, support vector machine) with ensemble algorithms (random forest, LightGBM, XGBoost, AdaBoost, Dagging, and Cascade Generalization) to identify the most significant factors influencing landslides. Using 228 landslide inventory maps, the study employed a random forest classifier and correlation-based feature selection (CFS) to determine the twelve most important parameters, including slope, elevation, aspect, geological features, and proximity to faults, roads, and streams. Slope was identified as the primary factor, followed by aspect and rainfall. The models were validated with metrics such as AUC, ACC, and Kappa, with XGBoost showing the highest performance (AUC 0.907, ACC 0.927, K 0.620) and logistic regression (LR) as the most accurate baseline model (AUC 0.784, ACC 0.912, K 0.394). The results highlight the effectiveness of LSM in guiding risk-mitigation strategies in geohazard-prone regions. The study demonstrates that integrating machine learning models enhances the accuracy and reliability of LSM, providing valuable insights for disaster management and policy-making. The final LSM was stratified into five susceptibility levels: very low, low, moderate, high, and very high, based on the natural breaks technique. The findings underscore the importance of slope, aspect, and rainfall in landslide prediction and the potential of ensemble methods in improving predictive accuracy. The study contributes to the understanding of landslide susceptibility factors and the application of machine learning in geohazard risk assessment.