Deep neural network (DNN) modelling for prediction of the mode of delivery.

One of the factors that worry obstetricians the most is the method of delivery. In recent years, the rate of caesarean sections has steadily climbed and now exceeds the threshold advised by medical organizations. Obstetricians typically lack the tools they need to assess whether vaginal delivery or a caesarean delivery is more appropriate. In this work, we suggested a computerized decision-making process for deciding on the best birthing style. The data was collected from 101 pregnant subjects who were admitted to hospital in eastern India for delivery from January 2021 to September 2021.The data set had 101 instances & 11 variables. The response was a binary variable with "caesarean" & "vaginal" as the outputs. A deep neural network model (DNN) was developed by using train set with h2o package. The model was selected on the basis of AUC (Area under the Curve) & KS (Kolmogorov-Smirnov) score. The AUC, KS score for train set were 0.99, 0.98 respectively. The prediction error rates for caeseraen & vaginal classes in train data are 0.02 & 0.00 respectively. The results support the use of these algorithms in the creation of a clinical decision system to help gynaecologists choose the most appropriate delivery method.