Smart insemination protocols based on CHAID decision trees for precision reproductive management and improved prolificacy in Murciano-Granadina does

• CHAID trees identify key drivers of prolificacy in Murciano-Granadina does. • Birth type is the dominant factor shaping reproductive outcomes. • Age and eCG dosage interact to influence fertility and neonatal survival. • Data-driven insemination protocols enhance precision and efficiency. • Seasonal and semen-type effects refine reproductive management. This study aimed to use a smart, data-driven analytical framework based on algorithmic rule-learning methods to explain and optimize artificial insemination protocols in Murciano-Granadina does. A total of 10,818 artificial inseminations from 8,121 does (45.16 ± 20.66 months old), using semen from 55 bucks, were analysed using canonical discriminant analysis and CHAID decision trees. The CHAID analysis followed a hierarchical structure, with variables selected sequentially according to explanatory power. Birth type acted as the primary predictor, followed by parity and age at insemination as secondary predictors, while equine chorionic gonadotropin (eCG) dose, season, and semen type contributed at lower hierarchical levels. Overall kid survival was high, with 98.4% live births and 1.6% stillbirths, mainly in high-order litters. Birth type emerged as the dominant determinant of prolificacy and kid survival, followed by parity, age at insemination, and eCG dose. Seasonal effects and semen type refined outcomes, with spring inseminations and fresh semen associated with greater homogeneity. Protocols combining these conditions with moderate eCG doses (250–300 IU) maximized prolificacy while minimizing stillbirth risk. Among successful inseminations, twin births predominated (57.2%), followed by single (24.9%) and triple (16.1%), whereas large litters (≥4 kids) were rare (1.9%) and associated with increased neonatal mortality. A significant age × eCG interaction was observed, with reduced fertility and increased neonatal losses in older or repeatedly treated does. CHAID models showed high classification accuracy, with Press’s Q values exceeding chance expectations, supporting precision, age- and parity-specific insemination protocols that enhance reproductive efficiency while reducing unnecessary hormonal use.