CNFE-SE: a novel approach combining complex network-based feature engineering and stacked ensemble to predict the success of intrauterine insemination and ranking the features

Abstract Background Intrauterine Insemination (IUI) outcome prediction is a challenging issue which the assisted reproductive technology (ART) practitioners are dealing with. Predicting success or failure of IUI based on couples' features can assist physicians to make appropriate decision for suggesting couples not and/or continuing treatment them. Many previous studies have been focused predicting in vitro fertilization (IVF) and intracytoplasmic sperm injection (ICSI) using machine learning algorithms. But, best our knowledge, few IUI. The main aim this study propose an automatic classification feature scoring method predict intrauterine insemination ranking most significant features. Methods For purpose, novel approach combining complex network-based engineering stacked ensemble (CNFE-SE) proposed. Three networks extracted considering patients' data similarities. step performed networks. original set engineered fed proposed classify per cycle. Our retrospective 5-year undergoing Data collected from Reproductive Biomedicine Research Center, Royan Institute describing 11,255 cycles 8,360 couples. dataset includes demographic characteristics, historical about diseases, clinical diagnosis, plans prescribed drugs during cycles, semen quality, laboratory tests pregnancy outcome. Results Experimental results show that outperforms compared methods with Area under receiver operating characteristics curve (AUC) 0.84 ± 0.01, sensitivity 0.79 specificity 0.91 accuracy 0.85 0.01 Conclusions important predictors parameters (sperm motility concentration) as well female body mass index (BMI).