Machine Learning-Based Polycystic Ovary Syndrome Generative Modelling via Ensemble Learning and Neural Networks for Infertility Prediction

• Caleb A. ABORISADE

  Department of Physics and Science Laboratory Technology, Abiola Ajimobi Technical University, Ibadan, Oyo State, Nigeria

  Author

• Jide E.T. AKINSOLA

  Department of Computer Sciences, Abiola Ajimobi Technical University, Ibadan, Oyo State, Nigeria

  Author

• Ifeoluwa M. OLANIYI

  Department of Computer Sciences, Abiola Ajimobi Technical University, Ibadan, Oyo State, Nigeria

  Author

• Fathia O. ONIPEDE

  Department of Computer Sciences, Abiola Ajimobi Technical University, Ibadan, Oyo State, Nigeria

  Author

• Emmanuel A. OLAJUBU

  Department of Computer Science and Engineering, Obafemi Awolowo University Ile–Ife, Nigeria

  Author

• Ganiyu A. ADEROUNMU

  Department of Computer Science and Engineering, Obafemi Awolowo University, Ile–Ife, Nigeria

  Author

Higher health issues, such as diabetes and hypertension, are sacrosanct with Polycystic Ovary Syndrome (PCOS) and they greatly affect fertility in women. Therefore, irregular menstrual periods, acne, increased hair growth and several hormone-related disorders are prevalent in people with PCOS. The study employed variable distributions and correlations, and the experimental design made use of exploratory data analysis and heat map visualization. Principal Component Analysis (PCA) was used for feature selection to minimize dimensionality and pinpoint the most informative attributes. A thorough cross-validation evaluation that suggested models' generalizability, robustness, and performance was considered. To ensure a thorough evaluation of the PCOS diagnostic value, evaluation measures such as F1-score, precision, accuracy and recall were utilized on Random Forest (RF), Recurrent Neural Networks (RNN) and Support Vector Machines (SVM) algorithms to build the models. The performance results show that RF had the best accuracy result of 99.74% followed by SVM with 99.21% and RNN with the worst result of 65.09%. This means that RF had the highest accurate predictions of the total amount of input samples. SVM and RF had the same precision result of 1.00, which shows that the two models had no misclassification of the PCOS infertility outcomes. That is, both SVM and RF could correctly identify all the positive instances and all the negative instances. The higher the value of the F1-score, the more reliable the model’s predictability. RF based on the highest F1-score of 0.9956 can be used for PCOS infertility modelling. The study concludes that RF is the golden model due to its superior performance for building a PCOS infertility prediction generative model. The study, therefore, suggests the implementation of federated learning and other deep learning algorithms for scalable performance using the big data paradigm.

• FJET_11_3_3

Copyright (c) 2025 FUDMA Journal of Engineering and Technology

This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License.