A Computational Intelligence Approach for Ranking Risk Factors in Preterm Birth

The aim of this paper is to propose a filter, based on a multi-objective evolutionary algorithm, for attributes' ranking in the context of a data mining task. The behavior of this filter is analyzed for the problem of ranking risk factors in preterm birth. The results obtained by applying the proposed evolutionary approach are compared with rankings obtained by applying some classical attributes selection methods and a logistic regression procedure. The influence of the ranking on a supervised classification (based on a radial basis function neural network) is also analyzed and the results suggest that the evolutionary approach provides a good quality ranking.