Early <scp>PCOS</scp> Detection: A Comparative Analysis of Traditional and Ensemble Machine Learning Models With Advanced Feature Selection

ABSTRACT PCOS (polycystic ovary syndrome) is a common hormonal disorder that affects many women during their reproductive years. It is marked by hormonal imbalances, leading to ovarian cysts, and can result in health issues such as infertility, diabetes, and even heart problems. Diagnosing PCOS accurately and early can be challenging, as it requires specific medical expertise. However, spotting PCOS promptly allows individuals to follow medical recommendations, which can lead to healthier lifestyles. In this study, we examined a dataset consisting of 541 patient records to enhance the detection of PCOS using advanced machine learning techniques. We established a data preprocessing pipeline that rigorously addressed missing values and identified outliers, while also normalizing the data to ensure it was ready for input. For feature selection, we applied advanced techniques such as SelectKBest, Chi‐Square, and XGBoost. These methods helped us pinpoint the most predictive attributes, which improved the interpretability and efficiency of our models. Hyperparameter tuning was carefully performed through grid search and cross‐validation, ensuring that each model was optimized for the best prediction accuracy. Importantly, our research highlights how effective machine learning can be in predicting PCOS. The logistic regression and support vector machine model stood out with its remarkable accuracy of 99.7753%. Furthermore, we created a user‐friendly web application to facilitate smooth deployment and real‐time analysis. This provides healthcare professionals with a handy tool for identifying early risks related to PCOS. The web application features an intuitive interface where users can easily input clinical information and receive immediate risk assessments.