Diagnosis of PCOS in Adolescent Girls Using Traditional and Ensemble Machine Learning Methods

Abstract Polycystic ovary syndrome (PCOS) affects women of childbearing age. PCOS is a condition where women have irregular menses, obesity, hyperandrogenism, type 2 diabetes mellitus, cardiovascular diseases, etc. Awareness of PCOS has been one of the most important concerns for female fertility as PCOS patient has different risks associated with health, among the above-listed risks, the major risk is infertility. In this study, we developed a Model for predicting PCOS among adolescent girls. The questionnaire was prepared by considering different symptoms related to PCOS and applying machine-learning techniques to detect and raise awareness of PCOS. A total of 21 features were available then feature selection techniques 14 best features among them were selected, we used, SMOTE(Synthetic Minority Oversampling Techniques) with under sampling to solve the problem of class imbalance, and then traditional and Ensemble ML models were created, we compared output from both conventional and ensemble models, and select the best model among them. We split the dataset 80:20, and results showed that the ensemble technique did better than a traditional method, where Extra Trees had 91.70% and BRF achieved an accuracy of 90.30%.