<p>Fertigated staked tomato (<i>Solanum lycopersicon</i>) is a productive crop. The dosage of fertilizers has been tested but their integration into a nutrient diagnosis and recommendation crop model is still pending. The objective of this paper is to guide the fertilization of staked tomatoes by linking nutrient diagnosis and crop response curves. Managerial, edaphic, physiological and climatic variables were documented at experimental sites from 2006 to 2024 in southern Brazil. Explanatory variables were related to marketable yield using Catboost classification and regression models. The tissue concentration values were centered log ratio (<i>clr</i>) transformed before computing nutrient standards and making yield predictions. The models were accurate (R² &gt; 0.85). The <i>clr</i>-transformed tissue nutrients were accurately related to yield by the classification model to derive nutrient standards from high-yielding and nutritionally balanced specimens. Yield was accurately related to <i>clr</i> variables and the data available in the database by regression models to simulate site-specific response curves for N, P and K and confirmed a diagnosis of N and P imbalance at a site showing several significant (<i>p &lt; 0.05</i>) growth-limiting factors. This unique database allowed linking nutrient diagnosis to fertilizer requirements for staked tomatoes under subtropical conditions using accurate machine learning models.</p> Graphical Abstract <p></p>