Abstract
BACKGROUND: In vitro fertilization (IVF) is a useful assisted reproductive technology to achieve pregnancy in infertile couples. However, it is very important to optimize the success rate after IVF by controlling for its influencing factors. This study aims to classify successful deliveries after IVF according to couples' characteristics and available data on oocytes, sperm, and embryos using several classification methods. MATERIALS AND METHODS: This historical cohort study was conducted in a referral infertility centre located in Tehran, Iran. The patients' demographic and clinical variables for 6071 cycles during March 21, 2011 to March 20, 2014 were collected. We used six different machine learning approaches including support vector machine (SVM), extreme gradient boosting (XGBoost), logistic regression (LR), random forest (RF), naïve Bayes (NB), and linear discriminant analysis (LDA) to predict successful delivery. The results of the performed methods were compared using accuracy tools. RESULTS: The rate of successful delivery was 81.2% among 4930 cycles. The total accuracy of the results exposed RF had the best performance among the six approaches (ACC=0.81). Regarding the importance of variables, total number of embryos, number of injected oocytes, cause of infertility, female age, and polycystic ovary syndrome (PCOS) were the most important factors predicting successful delivery. CONCLUSION: A successful delivery following IVF in infertile individuals is considerably affected by the number of embryos, number of injected oocytes, cause of infertility, female age, and PCOS.