ABSTRACT
The aim of this study is to compare the performance of Deep Neural Networks (DNN) and Ensemble machine learning methods on the cesarean data set. The data set consists of 80 people’s age, delivery number, delivery time, blood pressure, heart problem and cesarean section decision. The DNN models created consist of five input and two output neurons. The Ensemble model consists of Artificial Neural Networks, XGBoost and NB individual classifiers. Two different methods were used to create models and compare their performance. In the first method, the data set was randomly divided into 70% training and 30% test data (DSA-1 and Ensemble-1 models). In the second method, 5-fold cross-verification was used (DSA-2 and Ensemble-2 models). Model performances were evaluated according to Recall (R), Specificity (S) and Accuracy (ACC) metrics. In both the methods, the best classification performance was obtained in DNN models. The performance metrics of the models are as follows: DNN-1 model R = 0.92, S = 0.82, ACC = 0.88; Ensemble-1 model R = 0.92, S = 0.73, ACC = 0.83; DNN-2 model R = 0.77, S = 0.83, ACC = 0.80; Ensemble-2 model R = 0.77, S = 0.60, ACC = 0.70. Both DNN-1 and Ensemble-1 models correctly classify those who should be cesarean by 92%.
