Classifiers with synthetic oversampling pre-process for In Vitro Fertilization predictions

Abstract

Assisted reproductive technology (ART) has been explored extensively to establish the predictive models as one of the automated tool to measure the for-IVF quality outcomes depending on the Age, Anti-Müllerian hormone (AMH), Right ovary (RO), Left Ovary (LO), Number of eggs, No of Insemination, no of fertilized, Egg quality attributes. Artificial Intelligence (AI) field has connectionism based Artificial Neural Networks and they are capable of supporting the classification or estimation needed. Research community has been working with shallow and deep learners as part of machine learning and contributing improved results by combining various models and trying different frameworks for similar types of problems. In this article supervised learning is applied for training and testing the IVF dataset with limitations such as small data size and imbalanced class distribution. It is achieved with the set of selected classifiers and appropriate filters in the pre-processing, a weighted ROC score 84% and accuracy 75%.