An Intelligent AI-Driven Decision Support Framework for Early PCOS Screening using Clinical and Biochemical Data

Reading about early detection of Polycystic Ovary Syndrome (PCOS) makes one feel very essential as it can avoid several of the long-term reproductive and metabolic issues. In this paper, we will present a decision-support framework based on AI and that filters PCOS in its early stages based on clinical and biochemical data. Our my own free Kaggle dataset was 541 patients and 43 features. Our pipeline combines preprocessing, SelectKBest a feature selection algorithm based on ANOVA, SMOTE has been deployed to overcome the issue of class imbalance and predictive models such as XGBoost, a deep neural network (DNN), and a stacked ensemble. We have good predictive results on our experiments: XGBoost recalls a ROC-AUC of 0.974, and the group has the highest accuracy of 0.931. In general, the framework appears to be valid and reproducible to aid in the screening of PCOS.