Abstract: The promotion and application of green production technologies among farmers are crucial for ensuring food safety, protecting the ecological environment, improving resource use efficiency, and maintaining stable economic returns. Integrated irrigation and fertilization, which synchronizes water and nutrient delivery to crops, is a typical example of such green production technology. Drawing on survey data from 442 citrus family farms in Jiangxi Province, China, this study employs an XGBoost model combined with SHAP analysis to systematically identify the determinants of technology adoption and compares its performance with that of traditional regression models. The findings show that the interpretable machine learning model based on ensemble learning performs markedly better in terms of predictive accuracy. Among all explanatory variables, government subsidy intensity, planting scale, and policy publicity are the three most important factors driving the adoption of integrated irrigation and fertilization, all with significantly positive effects. Overall, external factors exert a stronger influence on adoption behavior than internal characteristics, and the interaction between planting scale and government subsidies generates a stronger incentive effect. These results provide useful guidance for governments seeking to further promote integrated irrigation and fertilization technology.