Predicting Infant Mortality Rates Across Fifty States Based on the ANN and SVM Modeling Approaches

Significance of the study: The infant mortality rate is an undeniably powerful indicator of the overall well being in a population. The prediction of infant mortality rates provide an opportunity for developing preventive strategies to improve population health. (Dollfus, 1990).

Purpose: This may be the first study using the ANN and SVM that will attempt to identify the important variable contributing infant mortality in the United States.

Method: Secondary data collected by the US Department of Labor in all states from 2000-2006 was utilized and tested through the Artificial Neural Network (ANN) and Support Victor Machine (SVM) to find the relatively important variables. The explanatory variables that were entered into the models are: per capita income, unemployment rates, number of teen mothers, number persons with grade 9 through 12 grade education without diploma, etc.

Significant Findings: In keeping with the current research in the field and according to preliminary analysis if the data collected the teenage pregnancy, level of education and per capita income variables in that order were the most inferential predictors of infant mortality in the country.

Although the U.S. infant mortality rate has fallen steadily in recent decades, the nation still ranked 27th among industrialized countries in an analysis of HHS 2000 data. The implications of this study are far reaching. If the most important predictors of infant mortality in the United States can be identified, targeted interventions can be put in place and the efficacy of programs already in place can be improved.