Exploring Machine Learning Models for Efficient Polycystic Ovary Syndrome Diagnosis

Abstract

Polycystic Ovary Syndrome (PCOS) is a common endocrine illness affecting women of reproductive age. It is distinguished by irregular menstruation, hyperandrogenism, and polycystic ovarian morphology. To lessen the long-term health effects of PCOS, early diagnosis and treatment are essential. Because of their ability to analyses complicated datasets and detect trends that human observers may miss, machine learning (ML) models have emerged as promising techniques for assisting in PCOS diagnosis. In this study, we investigate multiple machine learning models for effective PCOS diagnosis, highlighting their merits, limits, and possible uses in clinical practice. We examine the current state of ML-based PCOS diagnosis, identify obstacles, and suggest future research areas. different existing Machine Learning (ML) techniques, namely, Support vector classifier (SVC), logistic regression, Random Forest (RF) classifier, Decision Tree (DT), K-Nearest Neighbor (KNN), XGBoost classifier, and Catboost classification algorithms. We also try to implement the ensemble of XGBoost with random forest algorithm. The performance of the classifier, F1 Score is evaluated on six selected features (Follicle numbers[L/R], weight gain, skin darkening, hair growth and PCOS[Y/N]) for higher correlation value, according to evaluation metrices and results, we have found that PCOS correctly predicted with XGBoost and Random-forest ensemble method.