Genetic Feature Subset Selection for Gender Classification: A Comparison Study

We consider the problem of gender classification from frontal facial images using genetic feature subsetselection. We argue that feature selection is an importantissue in gender classification and demonstrate that GeneticAlgorithms (GA) can select good subsets of features (i.e.,features that encode mostly gender information), reducingthe classification error. First, Principal Component Analysis (PCA) is used to represent each image as a feature vector(i.e.,eigen-features) in a low-dimensional space. GeneticAlgorithms (GAs) are then employed to select a subset offeatures from the low-dimensional representation by disregarding certain eigenvectors that do not seem to encode important gender information. Four different classifiers werecompared in this study using genetic feature subset selection: a Bayes classifier, a Neural Network (NN) classifier,a Support Vector Machine (SVM) classifier, and a classifierbased on Linear Discriminant Analysis (LDA). Our experimental results show a significant error rate reduction in allcases. The best performance was obtained using the SVMclassifier. Using only 8.4%of the features in the completeset, the SVM classifier achieved an error rate of 4.7% froman average error rate of 8.9% using manually selected features.