Comparative analysis of machine learning models on prediction of polycystic ovary syndrome

Polycystic Ovary Syndrome (PCOS) is a major health condition in women existing globally, typically found by hormonal disruption and reproductive difficulties. Proper diagnosis of the syndrome, preventing it from developing to a critical phase can help in minimising further consequences. The significant objective of our study is to find the most suitable machine learning model and to identify the predominant variable, affecting the syndrome. The accuracy of three machine learning models namely Support Vector Machine, K-Nearest Neighbor and Decision Tree were compared to choose the ideal model for the prediction. Using SHAP summary plot, we identified the most influencing variable for the cause of the syndrome. Also, to validate the model's performance measures like precision, F1-score, recall, specificity and ROC curve were employed. On the overall analysis, Support Vector Machine was found to be the absolute model for the data set, in all aspects, dominating other models and manifested its superior ability in prediction the syndrome at initial stages.