PCOS Diagnosis Prediction Using Random Forest and Clinical Health Indicators

Polycystic Ovary Syndrome (PCOS) is a common
endocrine condition that affects women of reproductive age,
often resulting in reproductive, metabolic, and psychological issues. Timely and precise diagnosis continues to be difficult owing to diverse symptoms and overlapping clinical features. This work introduces a machine learning strategy for predicting PCOS diagnosis using the Random Forest algorithm and essential clinical health factors. A publicly accessible dataset including 1000 patient records was used, including age, body
mass index, menstrual irregularity, testosterone level, and
antral follicle count as predictive variables. Data preparation
included normalization, management of binary characteristics,
and stratified train-test splitting to maintain class equilibrium.
The Random Forest model was developed to identify nonlinear
correlations and feature interactions present in clinical data.
Experimental findings indicate that the proposed model
attained an overall diagnostic accuracy of 99.70%, exhibiting
robust sensitivity and specificity, hence facilitating efficient
differentiation between PCOS and non-PCOS patients. Analysis
of feature significance indicated that testosterone levels and
antral follicle count were the most significant predictors,
followed by BMI and menstrual irregularity. The results
underscore the efficacy of ensemble learning methodologies in
facilitating clinical decision-making. This methodology
facilitates scalability, interpretability, and clinical decisionmaking assistance.