Abstract: Soil stratification is a critical foundation in engineering geological investigation. The alluvial plain was characterized by complex sedimentary environment, poor stratum continuity and well-developed transition zones. Different engineers may draw reasonable judgments that comply with specifications yet yield divergent conclusions based on experience, which would introduce subjective ambiguity in stratification results. A weighted K-means clustering method incorporating a hybrid weighting strategy was proposed to objectivize the stratification process and quantify decision-making. Based on the survey data from a plain reservoir in the lower Yellow River, the authors selected 8 independent feature indicators from 18 geotechnical test parameters, that is dry density (ρ_d), liquid limit (w_L), plastic limit (w_p), mean particle size (d₅₀), internal friction angle (φ), cohesion (c), compression coefficient (a_1-2), and sampling elevation (SE). After KMO-Bartlett suitability verification and Z-score standardization, a hybrid weighting strategy integrating the Analytic Hierarchy Process (AHP), expert survey, and Principal Component Analysis (PCA) was employed. The optimal subjective-objective weight ratio (6∶4) was determined through sensitivity analysis. A weighted Euclidean distance clustering model was subsequently constructed, and the optimal number of strata (K=6) were identified using the elbow method. Validation results demonstrated that the stratification results by this method exhibited strong agreement with comprehensive expert judgments (Kappa=0.836), and it could achieve fine identification of stratigraphic transition zones (86.7% accuracy), confirming the effectiveness of the hybrid weighting strategy. This research provided a data-driven, reproducible, and objective decision-support tool for soil stratification in alluvial plains, offering technical support for standardizing stratification schemes.