Evaluating machine learning model performance in predicting polycystic ovarian syndrome

Abstract

The increasing incidence of polycystic ovary syndrome (PCOS) caused significant study. Polycystic ovary syndrome (PCOS) is a hormonal disorder that disrupts the men-strual cycle and affects a significant proportion of women in reproductive age. When compared to healthy women, those who have been diagnosed with Polycystic Ovary Syndrome (PCOS) typically experience less than eight menstrual cycles each year. Significant repercussions, such as infertility and ovarian cyst development, can result from menstrual cycle irregularities. Symptoms of polycystic ovary syndrome (PCOS) include but are not limited to: infrequent menstrual periods, excess weight, skin hyperpigmentation, insulin resistance, and high blood pressure. Even after being diagnosed with Polycystic Ovary Syndrome (PCOS), many women still don't understand what it means. This means that many people are still not receiving the care they need. The current study examines this question by comparing several machine learning techniques for detecting polycystic ovarian syndrome (PCOS), a precursor to the aforementioned disease. To conduct a comparative analysis with a sample size of 1500 women, a survey instrument consisting of a patient questionnaire was developed. This study highlights the significance and utility of twelve possible points gained from the examination of nineteen medical and physiological test outcomes. Logistic regression, K-Nearest Neighbor (KNN), Gaussian Naive Bayes, Random Forest Classifier, and Support Vector Machine (SVM) are used to identify polycystic ovary syndrome (PCOS). With a 96% accuracy rate, the Radial SVM was selected as the most suited and efficient technique for predicting PCOS.