Machine Learning-Based Predictive Modeling for Polycystic Ovary Syndrome Diagnosis: A Multi -Algorithm Comparative Analysis

Polycystic Ovary Syndrome (PCOS) is a multi-faceted endocrine disorder affecting reproductive, metabolic, and hormonal health in women. Early and accurate diagnosis is essential for effective management, yet conventional diagnostic methods often struggle due to the heterogeneous nature of PCOS symptoms and the lack of standardized criteria. This study proposes a machine learning-based framework for PCOS classification, integrating clinical, demographic, and biochemical data to improve diagnostic accuracy and reliability. The methodology encompasses data preprocessing, feature selection, model training, hyperparameter optimization, and rigorous evaluation to ensure robustness. Seven machine learning models were assessed, including Logistic Regression, Decision Tree, Random Forest, K-Nearest Neighbors (KNN), Gradient Boosting (GBM), Neural Networks, and Support Vector Machine (SVM). Among these, the SVM with a Radial Basis Function (RBF) kernel demonstrated the highest performance, achieving 90% accuracy, an AVC-ROC score of 0.98, and superior generalization capabilities. Feature importance analysis identified BMI, LH/FSH ratio, and Anti-Miillerian Hormone (AMH) levels as key predictors. Comparative analysis revealed that while Random Forest and Gradient Boosting also exhibited strong predictive performance, SVM out-performed them in classification accuracy and robustness. These findings validate the effectiveness of machine learning in PCOS classification, offering a scalable and data-driven complement to traditional diagnostic approaches. Future research will focus on deep learning architectures, the inclusion of larger and more diverse datasets for improved generalizability, and explainable AI (XAI) techniques to enhance model interpretability. Additionally, A real time prediction scenario was constructed with the trained model with user defined inputs which demonstrate the models potential for use case in PCOS diagnosis.