INVESTIGATING THE WORKABILITY OF RANDOM FOREST IN POLYCYSTIC OVARY SYNDROME DETECTION: A PROPOSED FRAMEWORK

Abstract

Recently, there has been a significant increase in the quantity of research focusing on the use of machine learning techniques to develop a model for PCOS detection. Given that the illness frequently affects women who are of reproductive age and results in infertility, the reasoning for this development is understandable. PCOS symptoms can fluctuate over a patients’ lifetime, making it difficult to diagnose. Irregular menstrual cycles, acne, hirsutism, and alopecia are common symptom, but their severity and presence can vary widely among patients. This study proposes investigating the workability of Random Forest (RF) in Polycystic Ovary Syndrome detection. The study utilized five models of random forest based on these parameters; estimators, bootstrap, depth, maximum features, maximum leaf nodes which was examined during the experimentation to ascertain the viability of the model. Model 3 of the random forest outperformed other models with an accuracy of 87.96% at 300 estimators, a true bootstrap, zero depth, maximum features is square root, and zero maximum leaf nodes.