Predicting IVF outcomes using a logistic regression-ABC hybrid model: A proof-of-concept study on supplement...

Area of Science:

• Reproductive Medicine
• Artificial Intelligence
• Biostatistics

Background:

• Machine learning (ML) is increasingly used in assisted reproductive technologies (ART).
• Existing studies often use conventional algorithms with limited optimization.
• Predictive modeling in in vitro fertilization (IVF) can be enhanced with advanced ML frameworks.

Purpose of the Study:

• To investigate a hybrid Logistic Regression-Artificial Bee Colony (LR-ABC) framework for enhanced predictive performance in IVF outcomes.
• To assess the interpretability of the LR-ABC model in identifying associations between supplement use and IVF success.
• To compare the LR-ABC hybrid model's performance against conventional ML algorithms.

Main Methods:

• Retrospective analysis of 162 women undergoing IVF, using 21 preprocessed clinical, demographic, and supplement variables.
• Implementation of four ML algorithms (KNN, CT, SVM, RF) and their LR-ABC hybrid counterparts.
• Evaluation using 5-fold cross-validation, SMOTE for class imbalance, and LIME for interpretability.

Main Results:

• LR-ABC hybrid models consistently outperformed their baseline counterparts across all tested algorithms (e.g., Random Forest accuracy increased from 85.2% to 91.36%).
• LIME explanations highlighted the influence of omega-3, folic acid, and dietician support on individual IVF predictions.
• The study demonstrated the methodological potential of LR-ABC for improving prediction and interpretability in IVF research.

Conclusions:

• The LR-ABC hybrid model shows promise for enhancing predictive accuracy and interpretability in IVF research.
• Identified associations between supplements and IVF outcomes are hypothesis-generating and require further validation.
• Future research necessitates larger, multi-center datasets with detailed supplement and dietary information to confirm findings.