Related Experiment Videos
Application of Machine Learning Algorithms for Evaluating Predictors and Developing Diagnostic Models for Female
Anwesha Dey1, Sandipan Das1, Rinku Saha2
1Department of Life Science and Bioinformatics, Assam University, Silchar 788011, India.
Bioengineering (Basel, Switzerland)
|July 28, 2026
Summary
Machine learning models can predict female fertility, identifying key factors like AMH and metabolic health. Naïve Bayes and Logistic Regression showed the most reliable results for diagnosing infertility.
Area of Science:
- Biomedical research
- Reproductive health
- Machine learning applications
Background:
- Infertility affects 1 in 6 people globally, impacting 12.6-17.5% of reproductive-aged couples.
- Machine learning (ML) offers potential for personalized medicine and diagnostic tools in healthcare.
- Predictive models can analyze measurable variables to aid in disease treatment and diagnosis.
Purpose of the Study:
- To evaluate the effectiveness of seven ML algorithms in predicting female fertility.
- To identify key predictors influencing female fertility.
- To enhance the interpretability of ML models in reproductive health using SHAP analysis.
Main Methods:
- Selected 20 predictor variables after multicollinearity testing for a binary supervised classification task.
- Evaluated seven ML algorithms: Logistic Regression, Random Forest, Decision Tree, Support Vector Machine, Naïve Bayes, K-Nearest Neighbour, and XGBoost.
- Utilized SHapley Additive exPlanations (SHAP) for model interpretability.
Main Results:
- Anti-Müllerian hormone (AMH) identified as a key biomarker for diagnosing polycystic ovarian syndrome (PCOS) and assessing ovarian reserve.
- Negative associations found between female fertility and waist circumference, systolic blood pressure, poor ovarian reserve, and triglycerides.
- Naïve Bayes and Logistic Regression demonstrated the most reliable and generalizable performance among the tested models.
- SHAP analysis highlighted polyendocrine metabolic ovarian syndrome (PMOS), AMH, poor ovarian reserve, menstrual cycle irregularity, systolic blood pressure, BMI, fasting glucose, and triglycerides as crucial predictors.
Conclusions:
- ML models, particularly Naïve Bayes and Logistic Regression, show promise for predicting female fertility.
- Metabolic factors and ovarian reserve markers significantly influence female fertility.
- Further multi-center studies with larger populations are needed to enhance ML reliability in clinical decision-making for infertility.
