Neural Networks Algorithm to Inquire Previous Preeclampsia Factors in Women with Chronic Hypertension During Pregnancy in Childbirth Process

Preeclampsia and eclampsia is one of the complications of pregnancy caused directly by the pregnancy itself. The success of handling preeclampsia is determined by maternity compliance in pregnancy care. Pregnant women who do not have their pregnancies cause no detectable high-risk factors experienced during pregnancy. We investigated the detection of preeclampsia by using neural network and also investigated the importance of the Previous PE Case attribute on the classification results. Preeclampsia dataset obtained from Haji General Hospital Surabaya with 17 parameters that are considered to affect the risk characteristics of the occurrence of preeclampsia, including the Previous PE Case parameters are positive or negative. This study uses Neural Network to classify Preeclampsia data. We compare with other algorithms like Naive Bayes, K-Nearest Neighbors, Linear Regression, Logistic Regression and Support Vector Machine. This experiment showed that the neural network algorithm achieved the best accuracy with three validation tests, 92,46% split data, 10-folds cross validation 94,23% and LOO validation 96.66%. The same data is applied to the neural network after excluding information about the previous PE case, for the learning process. The result is the correct classification as high as 96.66% of cases of case preeclampsia using all parameters, in the test set. Predicted cases of preeclampsia, for the total results of the unknown verification test, is 90%. And if the information on the previous PE case is not used, the result will decrease significantly.