Hybrid Models for PCOS Prediction using Machine Learning Algorithms to Address Class Imbalance

PCOS is a significant reproductive syndrome that is prevalent among women globally. It is distinguished by hormonal imbalances, specifically an excess of male hormones and elevated androgen levels, which disrupt normal ovulation and lead to various complications such as infertility, miscarriage, and pregnancy-related issues. Furthermore, individuals with PCOS have a higher likelihood of experiencing insulin resistance and being susceptible to the onset of type 2 diabetes.This study aims to address the challenges that are associated with identifying and classifying instances of PCOS. By balancing the dataset using SMOTE and employing machine learning methods such as Logistic Regression, Random Forest, and AdaBoost the research aims to improve PCOS identification accuracy. Correlation analysis and Principal Component Analysis (PCA) will be employed for individual algorithm. By incorporating both correlation analysis and PCA for individual algorithm, the hybrid model aims to improve the accuracy and effectiveness of PCOS identification. In this study, among the classification algorithms used, Logistic Regression emerged as the most effective model compared with other models for PCOS identification, achieving an impressive accuracy score of 95.4%. The findings of this study indicate that incorporating correlation analysis into machine learning models leads to improved performance in predicting or classifying instances. Specifically, the Logistic Regression model demonstrates superior accuracy, precision, recall, and F1 score compared to the ML models incorporated with the PCA.