Prediction of Preterm Pregnancies using Soft Computing techniques Neural Networks and Gradient Descent Optimizer

Abstract

This paper gives a study of the major risk factors which lead to preterm delivery in women. Preterm birth is the leading cause of perinatal morbidity and mortality worldwide. For the prediction of preterm delivery, inputs such as the height of the mother (maternal height), gravida (number of pregnancies) and para (number of pregnancies which crossed minimum gestational age) are used. To train the model for prediction, soft computing techniques such as Softmax regression using Neural Networks and Gradient Descent Optimizer are used. The success rate of prediction obtained is 89.99% with a stepwise cost of 0.52 on average. Hence, this model proves as a reliable predictor to identify women with a high risk of preterm, so as to provide sufficient time to plan for required antenatal and clinical interventions during pregnancy.