Predicting the delivery type of pregnancy: a comparative study

Abstract

One of the most beautiful things in the world is the birth of a child. The advancement of technology gives a mother several options to deliver a child to the world. However, unexpected incidences at the time of delivery may cause her to change the delivery type of the baby. Therefore, it would be essential and advantageous if there is a method that can predict the best way to make this happen based on the information collected from the mother and the baby to be born. This study intends to examine the best model to predict the delivery type. This study was conducted based on the medical records available from January 2015 to December 2015 in the General Hospital at Ampara, Sri Lanka. Births of 1400 babies from the mentioned metropolis are analysed in this study. There are two popular delivery types of birth, namely, Vaginal and Cesarean. The maternal factors such as ethnicity, age, the number of pregnancies, the number of babies along with the infant's gender, weight, head circumference, height and shoulder-length were recorded. Several classification models, namely, logistic regression, decision trees, support vector machines and a naive-Bayes classifier, were used to find the best-suited model to predict the delivery type. Moreover, the most significant factors that affect the type of birth were identified. Results indicated that the maternal age, delivery time, infant’s weight and infant shoulder-length have a statistically significant association with the type of delivery. The logistic regression model was obtained by getting delivery type as the dependent variable and the model was in a better fit with 68.21%. In the decision tree, the accuracy of the model was 71.43 %. Also, the most significant factor was the delivery time in the decision tree. In Naïve Bayes Classifier, the accuracy was 67.68% and Support Vector Machines outcome suggested that the accuracy was 79.11%. Study outcomes suggest that Support Vector Machines can be used to predict the delivery type in higher accuracy.