A Non-Invasive Diagnostic Screening Framework for Polycystic Ovary Syndrome (PCOS) Using Menstrual History and Mental Health Biomarkers: A Machine Learning Approach

This dataset contains an optimized matrix of non-invasive variables categorized into three core domains: (i) menstrual biometrics (cycle regularity, length, and flow intensity), (ii) phenotypic clinical manifestations (presence of acne and hair thinning/alopecia), and (iii) psychological affective distress vectors (standardized scores for stress, anxiety, and depression), all mapped against a binary target (PCOS diagnosis). Researchers and data scientists can utilize this repository to benchmark advanced classification architectures (such as XGBoost, Random Forest, and Deep Neural Networks), execute non-parametric Bootstrap LASSO or Boruta feature stability selections, analyze feature attributions via explainable AI (TreeSHAP) frameworks, and model probability calibration curves for low-resource telehealth screening triage.