A Comparative Analysis of Machine Learning Algorithms to Predict Polycystic Ovarian Syndrome (PCOS)

Abstract

In health services, artificial intelligence is used for early diagnosis functions to manage huge quantities of medical studies with great accuracy and precision. PCOS is a hormonal illness that affects a woman's menstrual cycle when she reaches reproductive age. Women with PCOS have a menstrual cycle that lasts 21 days or longer. Women with PCOS might have less periods (less than eight in a year) or stop having feminine cycles inside and out. This can end in infertility including the appearance of cysts in the ovaries. Irregular menstruation cycles, weight gain, skin darkening, thinning hair on the scalp, diabetes, and high blood pressure are all symptoms of PCOS. It's better to have a diagnosis and treatment as quickly as possible. Assortment of indications and the presence of an assortment of gynecological disorders, PCOS is especially hard to analyze. The time and money spent on the many clinical testing and ovarian scans has become a burden for PCOS sufferers. To resolve this concern, this paper compares machine learning methods for the initial prediction of PCOS using an ideal and basic but promising clinical and metabolic parameter that serves as an early marker for the condition. To collect the data needed for this comparative analysis, a patient questionnaire of 280 women was conducted during doctor consultations and clinical examinations. Based on the significance of the 19 features from medical and physiological test results, 12 prospective points are listed. In the Jupyter Python IDE, PCOS is identified using various machine learning techniques such as logistic regression, K-Nearest Neighbor (KNN), Gaussian Naive Bayes, Random Forest Classifier, and Support Vector Machine (SVM). Random Forest Classifier (RFC) was discovered to be the most appropriate and effective methodology for PCOS prediction, with an accuracy of 100 percent.