Abstract: Under the dual pressures of intensified global climate change and prominent ecological risks in arid and semi-arid regions of Western China, enhancing the resilience of cultivated land systems has become a crucial issue for ensuring regional food security and agricultural sustainable development. As a major production base for high-quality cotton and grain in China and an important ecological barrier, Xinjiang's oasis cultivated land faces multiple disturbances such as drought, wind-blown sand, and urbanization, resulting in significant systemic vulnerability. Investigating the resilience pattern and driving mechanisms of cultivated land in this region holds important theoretical and practical value. Based on ecosystem resilience theory, this study deconstructs cultivated land resilience into three capacities: resistance, adaptation, and recovery. A differentiated evaluation system comprising 14 indicators is constructed, tailored to the unique characteristics of Xinjiang’s oasis cultivated land, including irrigation dependence, ecological sensitivity, and economic underdevelopment. Resistance capacity is measured by indicators such as the proportion of effective irrigated area, soil thickness, soil pH, and soil moisture content. Adaptation capacity is represented by crop planting diversity index, multiple cropping index, agricultural machinery use per land area, and rural electricity consumption. Recovery capacity is assessed by per capita grain security rate, grain output rate of cultivated land, landscape fragmentation index, and per capita agricultural productive investment. To avoid estimation bias, eight control variables are introduced, including elevation, annual average rainfall, annual average temperature, pesticide use per land area, agricultural film use intensity, grain crop sowing area ratio, rural per capita disposable income, and per capita GDP. The study employs the XGBoost-SHAP model combined with kernel density estimation and spatial autocorrelation analysis to examine the spatiotemporal patterns of cultivated land resilience at the county level in Xinjiang from 2000 to 2023. The XGBoost model demonstrates strong predictive performance, with training set R² values exceeding 0.92 for all resilience dimensions and test set R² values ranging from 0.785 to 0.919. SHAP analysis quantifies the contribution of each driving factor and reveals nonlinear marginal effects. The results indicate that: (1) During the study period, the overall resilience of cultivated land in Xinjiang showed a fluctuating upward trend, with the peak of the kernel density curve shifting to the right and the width narrowing, indicating a reduction in regional disparities. (2) Spatially, a “high in the north, low in the south” pattern was observed, with high-value areas concentrated in eastern oases (e.g., Yizhou District, Qitai County) and gradually clustering, while the extent of low-value areas shrank. The global Moran’s I index ranged between 0.182 and 0.310, indicating significant spatial aggregation. (3) Rainfall, per capita GDP, and elevation were identified as core drivers influencing cultivated land resilience, with contribution shares of 36.5%, 16.7%, and 13.9%, respectively. Increased rainfall may weaken resilience due to waterlogging or soil nutrient loss, while mid-elevation areas exhibited higher resilience. Reasonable investment in per capita GDP can enhance cultivated land resilience, though excessive investment may lead to diminishing returns. The findings provide scientific support for formulating differentiated cultivated land protection strategies in Xinjiang and similar arid and semi-arid regions worldwide. The study also demonstrates the effectiveness of interpretable machine learning models in uncovering nonlinear driving mechanisms of cultivated land resilience, offering a methodological reference for future research.