The abstract Common fertilizer-scheduling methods often propose a singular solution for nutrient administration, which may not be best possible in diverse farming contexts. This type of research presents a comprehensible XGBoost-based method that incorporates soil uniqueness, meteorological factors, and crop specific information to generate accurate compost recommendations. The innovative idea here is to apply SHAP (SHapley Additive exPlanations) for replica interpretability, lets farmers see the most important factors that led to the recommendation. The statistics include 1,200 trials of NPK standards, pH, temperature, rainfall, top soil colour, and harvest type. These samples be pre-processed using consistency and ticket programming. With 92.3% accuracy on experimental data, grid-search-based overexcited factor change outperformed arbitrary forest and SVMs, which had accuracy rates of 88.5% and 84.7%, respectively, in replica optimization. The two most significant factors identified by SHAP research are rainfall (weight = 0.18) and nitrogen (weight = 0.35) in the top soil. The usefulness of this structure was demonstrated via a user interface that provided recommendations in real-world situations. The position demonstrates a link between farming ingenuity and a superior device teaching strategy.