Enhancing PCOS Prediction: A System based on Ensemble Machine Learning Techniques
M. Abdul Goffar Khan, Farzana Nila, Nafisa Tabasaum, Sayma Alam Suha, Muhammad Nazrul Islam
Abstract
Polycystic Ovarian Syndrome (PCOS) has become the most common hormonal disease in women who are in their reproductive age. PCOS in women affects the reproductive organs that generate the hormones progesterone and estrogen, which regulate the menstrual cycle. Although there are not that many symptoms in the early stage of PCOS, as the condition continues, they become much more obvious, and at that point, the treatment becomes much more difficult. Again, the Machine Learning (ML) model can help to predict PCOS at the early stage. Therefore, the main objective of this research is to develop an ensemble-based machine learning model for predicting PCOS and to develop a web system based on the ensemble model for predicting PCOS. To achieve these objectives, best set of ML algorithms were selected by exploring performances (accuracy and false negative error) of ten ML algorithms were explored. An ensemble stacking classifier was proposed by integrating the six classifiers that showed an accuracy of 93.74% for predicting the PCOS. A web system was also developed adopting the proposed ensemble model for predicting PCOS by the end users.