Enhancing the Healthcare by an Automated Detection Method for PCOS Using Supervised Machine Learning Algorithm

Identifying Poly Cystic Ovary Syndrome (PCOS) can be challenging due to several factors like wide spectrum of symptoms, that varies according to individual, overlapping symptoms with other conditions like thyroid disorders, adrenal gland disorders, and ovarian tumors. Many women tend to overlook the common indications of PCOS and only seek medical attention when they encounter difficulties in conceiving. The integration of Machine Learning (ML) tools into clinical practice in collaboration with healthcare professionals have guaranteed proper interpretation and clinical decision-making. In this research, we explored the feasibility of developing a model that utilizes ML algorithms and techniques to automate the diagnosis of PCOS. To accomplish this, we utilized a dataset containing information from 543 subjects, encompassing 42 features such as metabolic, imaging, hormonal, and biochemical parameters. Initially, data pre-processing is performed, followed by the implementation of feature selection approach to reduce the number of features. Subsequently, various classification algorithms were trained and evaluated using Decision Tree (DT), XG Boost, Ada Boost, Random Forest (RF) and Logistic Regression (LR). After conducting a comprehensive analysis, we determined that the XG Boost outperformed the other algorithms in terms achieving high accuracy 92.45%.