Flood susceptibility provides scientific support for flood prevention planning and infrastructure development by identifying and assessing flood-prone areas. The uncertainty posed by non-flood sample datasets remains a key challenge in flood susceptibility mapping. Therefore, this study proposes a novel sampling method for non-flood points. A flood susceptibility model is constructed using a machine learning algorithm to examine the uncertainty in flood susceptibility due to non-flood point selection. The influencing factors of flood susceptibility are analyzed through interpretable models. Compared to non-flood datasets generated by random sampling with the buffer method, the non-flood dataset constructed using the spatial range identified by the frequency ratio model and sampling method of one-class support vector machine achieves higher accuracy. This significantly improves the simulation accuracy of the flood susceptibility model, with an accuracy increase of 24% in the ENSEMBLE model. (2) In constructing the flood susceptibility model using the optimal non-flood dataset, the ENSEMBLE learning algorithm demonstrates higher accuracy than other machine learning methods, with an AUC of 0.95. (3) The northern and southeastern regions of the Zijiang River Basin have extremely high flood susceptibility. Elevation and drainage density are identified as key factors causing high flood susceptibility in these areas, whereas the southwestern region exhibits low flood susceptibility due to higher elevation. (4) Elevation, slope, and drainage density are the three most important factors affecting flood susceptibility. Lower values of elevation and slope and higher drainage density correlate with higher flood susceptibility. This study offers a new approach to reducing uncertainty in flood susceptibility and provides technical support for flood prevention and disaster mitigation in the basin.

Flood susceptibility mapping in the Yellow River Basin remains challenging due to uncertainties in sample selection and model generalization. This study develops a novel two-step coupling framework that integrates an information value (IV) model with machine learning (ML) to improve reliability. The IV model first identifies stable low-susceptibility zones to select robust non-flood samples, which are then combined with historical flood inventories to train ML models. The SHAP method is applied to quantify factor contributions and interpret outputs. The results show that the IV-RF model achieves the highest predictive performance, while a stacking ensemble further reduces uncertainty. Sensitivity analyses confirm that model outcomes remain stable across random data splits and repeated non-flood point selections. High-risk areas are primarily located in the Hetao Plain, the Weihe River Basin, and sections of the lower Yellow River Basin, where susceptibility is driven by drainage density, anthropogenic and urban–mining soils, a high topographic wetness index, and gentle slopes. This work provides a transferable methodology that enhances physically consistent sample selection and model interpretability for flood risk assessment.