LEVERAGING MACHINE LEARNING MODELS FOR PREDICTING THE LIKELIHOOD OF POLYCYSTIC OVARIAN SYNDROME IN WOMEN OF REPRODUCTIVE AGE

Abstract

Conventional diagnostic approaches for polycystic ovarian syndrome (PCOS – a condition characterized by heterogeneity and the absence of a singular diagnostic test) are often invasive, time-consuming, and rely on varying criteria, resulting in inconsistencies in diagnosis. This study addresses the pressing challenge of improving the diagnosis of PCOS by exploring machine learning applications to bridge gaps in its prediction and diagnosis, offering a potential pathway toward greater accuracy and efficiency. The Cross-Industry Standard Process for Data Mining methodology was adopted for implementation using a comprehensive dataset from a public library – Kaggle. Results identified XGBoost algorithm as the most effective predictive model for diagnosing and predicting PCOS, achieving an accuracy of 98.7%. The results of the study indicated that the XGBoost algorithm is reliable with strong accuracy and dependability in diagnosing PCOS, establishing the PCOS Predictor as a valuable tool in clinical environments.  This study thus represents a significant step forward in transforming the diagnostic landscape of PCOS, combining technological advancements with clinical insights to enhance women's healthcare.