Ensemble Machine Learning Framework for Early Diagnosis of PCOS/PCOD and Infertility Risk Prediction

Abstract

Polycystic Ovary Syndrome (PCOS) and Polycystic Ovarian Disease (PCOD) are prevalent endocrine illnesses impacting women of reproductive age, frequently linked to hormonal imbalances, metabolic issues and infertility. The overlapping clinical symptoms and the constraints of conventional diagnostic techniques render the early diagnosis of PCOS or PCOD a considerable clinical problem. This paper provides a machine learning framework for precise illness diagnosis and infertility risk prediction. The methodology includes a systematic pre-processing phase to remove extraneous and irrelevant data, thereby enhancing data quality and model efficacy. Recursive Feature Elimination (RFE) is utilised to determine the most pertinent features for classification. Three supervised machine learning models C4.5 Decision Trees, Bagged Decision Trees and Boosted Decision Trees are assessed utilising PCOS Prediction datasets. Experimental findings indicate that ensemble learning methods, specifically bagged and boosted decision trees, surpass solitary decision tree models in classification and prediction tasks. The proposed Ensemble Learning model attained an accuracy of 96.03%, demonstrating its efficacy in detecting PCOS/PCOD cases and evaluating infertility risk. These findings indicate that the use of comprehensive pre-processing, feature selection and ensemble learning methodologies can markedly improve early diagnosis and facilitate prompt clinical intervention, hence enhancing reproductive health outcomes.