Prediction of Newborn Weight Using Questionnaire Data and Machine Learning Approach

Abstract

Low birth weight is an important factor of the health of an infant because of being a major determinant of morbidity, disability, and mortality during infancy and childhood. Low birth weight is a multifaceted public health problem. The prevalence of this diagnosis is estimated to be 15–20 % worldwide (more than 20 million infants) and occurs mostly in developing countries (95.6 %). There are many factors, which affect the birth weight of newborns. From a medical point of view, it is important to find out, which factors are important and determine the risk of not normal birth weight. In this work, we propose to use the classification system to find out which features are important to classify the neonates into birth weight groups. In presented approach, we use different classification and feature selection methods to analyze a new questionnaire-based obstetric data set. Our results show that a small number of features is enough to achieve high classification accuracy rate and the set of selected features is significant from biological and medical point of view.