Methodology for a Dual-Target Reproductive Intelligence System for Early Prediction of Infertility and Menopause

Reproductive health of women involves complex, heterogeneous and interdependent clinical factors that make early risk assessment difficult through manual valuation alone. This study proposes a dual-target machine-learning reproductive intelligence system for predicting infertility risk and menopause transition from shared health data. The system integrates clinical records, hormonal profiles, laboratory results, ultrasound findings, menstrual history, age, body mass index, lifestyle factors, infection history and patient-reported symptoms. It addresses a limitation in existing reproductive-health applications which commonly focus on isolated tasks, such as ovulation tracking, embryo grading, assisted-reproduction outcomes, symptom monitoring, without jointly modelling infertility and menopause across the reproductive life course. The proposed architecture uses a stacked ensemble of Extreme Gradient Boosting, Extra Trees and Convolutional Neural Networks as base learners. Their predictions are combined via a Random Forest meta-learner to classify women into infertility-risk categories and menopause-transition stages. This design exploits complementary strengths in nonlinear feature learning, variable interaction detection and robust classification. SHAP additive explanations are incorporated to identify the contribution of each predictor and provide patientlevel and global explanations. Thereby, improving transparency and clinical interpretability. The system is intended to support earlier screening, risk stratification, referral, and personalised reproductive-health decisions rather than replace professional diagnosis. Its application is especially relevant in Nigeria and similar African settings where reproductive data remain underused and delayed assessment, repeated hospital visits, stigma, emotional distress and limited menopause services persist. The proposed methodology provides an integrated, explainable and context-sensitive foundation for intelligent decision support across the reproductive ageing continuum in women