Influence of Feature Selection on Naïve Bayes Classifier for Recognizing Patterns in Cardiotocograms

Cardiotocography is a technical procedure that consists in recording the fetal heart rate (FHR) and uterine activity (UA) during the last months of a pregnancy. Cardiotocogram (CTG) analysis consists in identifying some patterns associated to fetal activity in order to detect potential fetal pathologies. Several automatic classification methods have been already tested on CTG data sets, while a few feature selection (FS) methods have been considered. The aim of this paper is to investigate the influence of FS on the performance of a naïve Bayes classifier for FHR patterns and fetal states. We empirically compare the performance of several models using four different FS methods (Correlation-based, ReliefF, Information Gain, and Mutual Information). We find that ReliefF yields to a better performance for fetal state classification, while no FS method worth the effort for FHR pattern classification. 