EMPOWERING PCOS DIAGNOSIS: A MACHINE LEARNING DECISION TREE APPROACH

Abstract

Polycystic Ovary Syndrome (PCOS) is a prevalent endocrine disorder that affects women of reproductive age and is often difficult to diagnose due to variability in clinical symptoms. This study presents an interpretable machine learning approach for PCOS diagnosis using a Decision Tree classifier trained on clinical and hormonal data. The dataset was obtained from Kaggle and consists of anonymized records of 200 women, including both PCOS and non-PCOS cases. Key features such as age, body mass index (BMI), menstrual irregularity, testosterone levels, and antral follicle count were used for model development. The proposed model achieved an overall accuracy of 84% and a recall of 75% for PCOS cases, demonstrating its effectiveness in identifying individuals at risk. Feature importance analysis highlighted menstrual irregularity, testosterone levels, and BMI as the most influential predictors, consistent with established clinical findings. However, the study is limited by the use of a single publicly available dataset and the lack of external validation. The results indicate that interpretable machine learning models can support early PCOS screening and clinical decision-making.