Abstract: Landslides rank among the most frequently occurring geological hazards worldwide. Identifying the dominant factors controlling landslide occurrence holds prime significance for mitigating the resulting losses. However, previous studies have not yet reached a unified understanding of such factors. Hence, this study investigated the Fushun area, where landslides are particularly frequent. The slope aspect, lithology, and faults were found to play a decisive role in the initiation of landslides in the area. In this study, nine evaluation factors, including the slope gradient and lithology, were selected and numerically normalized using the weight of evidence method. The factor weights and final susceptibility maps were obtained using the random forest (RF) and multi-layer perceptron (MLP) models. The results were compared and validated through the receiver operating characteristic (ROC) curve analysis to ensure reliability. The validation results indicate that all models achieved accuracies of above 80%, revealing high proportions of landslides in very high- and high-susceptibility areas. Therefore, the hybrid machine learning models were reliable and applicable. Notably, the MLP-based hybrid models performed the best, attaining an accuracy of 83.1%. The factor weight analysis demonstrates that the southbound slope aspect, semi-weak rocks, and faults represent the dominant contributors to landslide formation in the Fushun area. Overall, these findings provide a reference for analyzing the dominant factors controlling landslide occurrence in similar areas, as well as reliable data for targeted disaster prevention and mitigation in the Fushun area.