Embryo Quality and in Vitro Fertilization Outcome Prediction Using Machine Learning and Explainability

Evaluation of embryo quality is crucial for the success of implantation and the rate of live birth in assisted reproductive technologies. Traditional morphological assessments are limited to the observer's experience. This study explores the application of various machine learning techniques, such as Random Forest, XGBoost, and Logistic Regression, to predict embryo viability and in vitro fertilization outcomes. A set of preprocessing activities was performed to handle missing embryological characteristics, encode categorical variables, and transform complex data formats into usable characteristics in the dataset. The best model was Random Forest which demonstrated robust predictive capabilities achieving a precision of 98.71 % and the lowest false positive rate with $\mathbf{0. 5 8 \%}$. The Shapley Additive Explanations algorithm allowed to identify key predictors such as embryo quality scores, patient age, and BMI enhancing explainability, providing a balance between model performance and clinical interpretability. These findings highlight the transformative potential of machine learning in reproductive medicine by providing precise, transparent, and scalable tools for embryo evaluation and prediction of In Vitro Fertilization-IVF outcomes.