Comparative Analysis of Predictive Models for the Likelihood of Infertility in Women Using Supervised Machine Learning Techniques

Abstract

Infertility is a worldwide problem, affecting 8% – 15% of the couples in their reproductive age. WHO estimates that there are 60 - 80 million infertile couples worldwide with the highest incidence in some regions of Sub-Saharan Africa also infertility rate may reach 50% compared to 20% in Eastern Mediterranean Region and 11% in the developed world. Infertility has caused considerable social, emotional and psychological stress between couples, among families, within the individual concerned and the society at large. Historical data constituting information describing the risk factors of infertility alongside the respective infertility likelihood status of women was collected from Obafemi Awolowo University Teaching Hospital Complex (OAUTHC).  The predictive model was formulated using naive Bayes’, decision trees and multi-layer perceptron algorithm – supervised machine learning algorithms.  The formulated model was simulated using the Waikato Environment for Knowledge Analysis (WEKA) environment.  The results of the performance evaluation of the machine learning algorithms showed that the C4.5 decision trees and the multi-layer perceptron with an accuracy of 74.4% each outperformed the naive Bayes’ algorithm.  In addition, the decision trees algorithm recognized variables relevant to predicting infertility and a rule that can be applied on patient risk factor records for infertility likelihood prediction was deduced from the tree structure.  This showed how effective machine learning algorithms can be used in predicting the likelihood of infertility in Nigerian women.