Diagnosis of Polycystic Ovary Syndrome Using Feature Selection-Based Machine Learning Algorithms

This study delves into the realm of employing cutting-edge machine learning algorithms to ascertain the diagnosis of polycystic ovary syndrome (PCOS) in women, utilizing a meticulously curated dataset obtained from the renowned Kaggle repository. This research utilizes a carefully curated dataset from the renowned Kaggle repository, which contains detailed information on 541 women, 177 of whom have been diagnosed with PCOS. The first step in this study involves employing a meticulous univariate feature selection algorithm. Its purpose is to uncover the most crucial predictors for PCOS within the dataset. Surprisingly, the analysis reveals that the ratio of Follicle-stimulating hormone (FSH) to Luteinizing hormone (LH) stands out as the most significant attribute when it comes to predicting PCOS. After an extensive analysis, the study narrows its focus to the top 10 features that are most strongly associated with PCOS. This condensed subset of attributes proves to be highly effective in accurately predicting the presence of PCOS. The outcome of this study is the outstanding performance of the RFLR model. Through a robust 40-fold cross-validation approach, this hybrid model achieves a remarkable testing accuracy of 91.01% and an impressive recall value of 90%. In other words, it effectively distinguishes and classifies PCOS patients with a high level of accuracy and sensitivity.