A Comprehensive Study of Machine Learning Techniques for Polycystic Ovary Syndrome Diagnosis

Polycystic Ovary Syndrome (PCOS) is the most common disorders which significantly effects on the quality of women life. Early diagnosis is crucial to effectively and appropriately manage and treat it. In this work, we proposed novel approach with advanced machine-learning techniques for predicting the possibility of having PCOS, based on clinical statistical data. For this the model is trained on Kaggle dataset containing 541 instances with two classes, presence and absence of PCOS. First, we pre-processed the dataset to get quality features. And applied different algorithms like Support Vector Machines, XGBoost, Artificial Neural Networks (ANN), and Random Forests, to predict performances with various methodologies. Every model was evaluated with accuracy, precision, and recall metrics to deduce a good predictive approach for diagnosing PCOS. Out of all methods the XGboost, RF models performed well.