The Loess Plateau, characterized by intense neotectonic activity, frequent earthquakes, and the unique physical and mechanical properties of loess (granular overhead pore structures), is prone to seismic-induced landslides, causing substantial casualties and economic losses. This study evaluates seismic landslide hazards in Tianshui City by coupling statistical and machine learning models. Six coupled models were developed: frequency ratio (FR) combined with logistic regression (LR), support vector machine (SVM), multilayer perceptron (MLP), and information value (IV) combined with LR, SVM, and MLP. Model performance was assessed using the area under the receiver operating characteristic curve (AUC), accuracy, precision, specificity, sensitivity, and F 1-score. The results show (1) as a branch of machine learning, the deep learning model (MLP) outperformed traditional machine learning models (LR and SVM), with FR-MLP and IV-MLP achieving higher AUC values than FR-LR/FR-SVM and IV-LR/IV-SVM,respectively. (2) IV-based models demonstrated superior performance over FR-based models, with the following AUC trends: FR-LR ＜IV-LR, FR-SVM ＜IV-SVM, FR-MLP ＜IV-MLP.(3) The IV-MLP model outperformed the other five models in all evaluation metrics, achieving an AUC of 0.93. These findings enrich methodologies for assessing seismic loess landslide hazards and provide critical insights for disaster prevention strategies.