IVF Success Rates Prediction Using Hybrid ANN-GA based Machine Learning Model

Abstract

Machine learning techniques have been studied with the aim of improving the predictions of In Vitro fertilization (IVF) Live Birth occurrence rates by assessing the extrinsic and intrinsic parameters that principally influence IVF process. Predictive performance of the machine learning techniques is directly related to the quality of the training database and also on the set of hyperparameters screened in the prediction process. Obtaining the best hyperparameters is not a trivial task, but can be achieved by implementing bioinspired algorithms such as Artificial neural network (ANN) and Genetic Algorithms (GA). ANN-GA hybrid design works based on the natural selection theory and evolve the solutions that produce good hyperparameters for Machine learning techniques to register higher accuracy predations Predictions. The IVF/ANN-GA has the aim to improve the performance of hybrid machine learning design with the addition of IVF-Inspired mechanisms that better exploit the information of individuals. With this aim, the present study explores the combination of an ANN with GA to search for the best set of hyperparameters to predict the success rates of the process. The results supported with high accuracy, precision, and recall. Performance values of the model such as F1-measure precision 0.85, recall values 0.76, F1_score 0.80 and accuracy measure 0.89 were noted. The measured values indicate that the model applied exhibits the true positive detection rate of 85%. Models detecting with false positives chance is measured to be only 15%. Study concludes that, present investigation rely both on precision and recall and which were successfully considered in the study metrics. F1 score of the employed design explains the arithmetic ratio of both precision and recall with 89% value. The present studied ANN-GA hybrid model achieved the overall accuracy rates of 90% in predicting the IVF Live Birth rates measures.