Abstract

Support Vector Machine (SVM) and Artificial Neural Network (ANN) approaches have been widely used for pattern recognition and data prediction. These approaches can be used to establish nonlinear relationships among variables as well as to perform variable ranking and selection. The purpose of this study was to construct workable SVM and ANN models to assess risk factors affecting infant mortality rates in the United States (U.S.). Data was collected in years 2000-2006 at the state level from the U.S. Census Bureau, the Center for Disease Control and Prevention, and the U.S. Departments of Labor and Commerce. The outcome variable was infant mortality rate, and risk factors included per capita income, unemployment rate, teen pregnancy rate, percent of teen mothers who smoke, percent of newborns with gestation stage less than 37 weeks, percent of newborns who weigh less than 2500 grams, and many more. The top seven important risk factors found in the SVM and ANN models were consistent with the literature findings, which are major contributors of the infant mortality rate. It is evident that SVM and ANN models have successfully demonstrated their model validity and applicability. Therefore, the study results should be used to help policy makers implement intervention strategies needed to reduce infant mortality rate.