Abstract: Precision water and fertilizer management is an important approach for achieving high-yield and high-efficiency winter wheat production. However, the nonlinear interactions among water, fertilizer, soil, and climatic factors under complex field conditions, as well as their key regulatory thresholds, remain insufficiently understood. This study systematically compiled field experimental data on winter wheat from published literature in Shandong Province during 2001-2024. After data screening, missing value processing, and outlier removal, 433 valid treatment observations were obtained to construct machine learning models for identifying key yield drivers and water-fertilizer management pathways. A random forest (RF) model was used to identify the major drivers influencing winter wheat yield, while a classification and regression tree (CART) model was applied to extract agronomically interpretable yield differentiation thresholds. The results showed that: (1) irrigation amount, nitrogen application rate, and soil available phosphorus exhibited high predictive importance and were key variables explaining winter wheat yield variation in Shandong Province; (2) soil alkalinity was an important constraint on yield potential. When soil pH ≥ 8.19, the mean yield of samples decreased significantly. Under conditions of high soil bulk density (≥1.31 g/cm³), higher growing-season precipitation (≥182.75 mm) was associated with lower yield levels, indicating that soil chemical properties and physical structure jointly affected yield formation; (3) the high-yield differentiation pathways identified by the CART model showed that, within the dataset analyzed, high-yield samples exceeding 10,000 kg/hm² mainly occurred under combinations of relatively high nitrogen application rates (≥225.00 kg/hm²) and relatively low phosphorus application rates (<112.50 kg/hm²). The integrated RF and CART approach effectively identified the key yield drivers and nonlinear differentiation pathways of winter wheat, providing theoretical support for optimizing regional water and fertilizer management and improving precision management strategies. It should be noted that the water and fertilizer thresholds identified in this study represent empirical breakpoints derived from an integrated literature-based dataset, and their applicability requires further validation through multi-year and multi-site field experiments.