Cattle discard is a critical decision-making process in dairy production, directly affecting herd productivity, resource efficiency, and farm profitability. Traditionally, these decisions have relied on the subjective judgment of farm managers, often lacking precision and consistency. This study addresses the limitations of manual discard practices by developing an intelligent classification model that utilizes supervised machine learning techniques to support data-driven decision-making. The proposed model was constructed using the CRISP-ML methodology, which structured the workflow from data acquisition to model deployment. Historical data spanning 1990 to 2019 from a dairy farm in Ecuador were analyzed, including variables related to milk production, reproductive history, animal health, and physiological traits. Three machine learning algorithms were evaluated: Random Forest, Support Vector Machines, and Logistic Regression. Among these, the Random Forest classifier demonstrated the best performance, achieving an accuracy of 96.47%, an F1 score of 96.41%, a recall of 96.47%, and an ROC-AUC of 98.31%. Feature importance analysis revealed milk yield, disease frequency, age, and reproductive indicators as the most influential variables in discarding decisions. The study presents a robust, interpretable, and deployable model that significantly enhances the precision of cattle discard classification. Its practical implementation in dairy farms can optimize herd composition, reduce operational costs, and support sustainable livestock management by promoting evidence-based decision-making. This study contributes to the advancement of artificial intelligence in agriculture, aligning with the goals of precision farming and the digital transformation of the livestock sector.