Predicting low birth weight babies through data mining

Abstract

Low Birth Weight (LBW) babies have a high risk of developing certain health conditions throughout their lives that affect negatively their quality of life. Therefore, a Decision Support System (DSS) that predicts whether a baby will be born with LBW would be of great interest. In this study, six different Data Mining (DM) algorithms are tested for five different scenarios. The scenarios combine information about the mother’s physical characteristics and habits, and the gestation. Results are promising and the best model achieved a sensitivity of 91,4% and a specificity of 99%. Good results were also achieved without considering the gestational age, which showed that the use of DM might be a good alternative to the traditional medical imaging exams in the prediction of LBW early in the pregnancy.