Explainable AI-based Predictive Model for Early Diagnosis of Polycystic Ovary Syndrome (PCOS)

Polycystic Ovary Syndrome (PCOS) is a hormonal disorder, which typically is common in women of reproductive age. The numerous symptoms of it result in a large number of cases being undiagnosed. Given that long-term effects that can be associated to PCOS such as infertility, obesity and diabetes can be minimized by early diagnosis, there can be no important role of PCOS diagnosis. The current paper presented an AIbased predictive model which can be used to identify PCOS at an early level of development with the assistance of clinical and lifestyle data. The data set was ready keeping a keen eye on filling in the lost values, normalization of the numerical attributes and coding of the categorical variables. Several machine learning algorithms, such as Logistic Regression, Support Vector machine, Naive Bayes, Decision Tree and Random Forest, were trained and compared by their performance metrics, which include-accuracy, precision, recall, F1-score and ROC-AUC. The most credible and precise outcomes were presented by the Random Forest model amongst them. In order to make the system more transparent and trustworthy, the Explainable AI (XAI) procedures of SHAP and permutation importance have been used to highlight the most significant effects in predictions. The attributes of BMI, follicle count, and length of menstrual cycles were proved to play a significant role in the development of risk of PCOS. In addition, a Gradio-based user interface was developed to enable real-time predictions and customized recommendations on lifestyle. Overall, the present AI system demonstrates that machine learning can be useful in the examination of early PCOS and aid in improving the health outcomes of women.