Applying GECs for Feature Selection and Weighting using X-TOOLSS

In [1], Abegaz et. al compared hybrid genetic and evolutionary feature selection (GEFeS) and weighting (GEFeW) on feature sets obtained by Eigenface, LBP, and oLBP feature extraction methods. GEFeS and GEFeW were implemented using a Steady-State Genetic Algorithm (SSGA). In this paper, we extend the work performed in [1] and compared GEFeS and GEFeW implementations using SSGAs and Estimation of Distribution Algorithms (EDAs). Our results show that GEFeS and GEFeW enhance the overall performance of both the Eigenfacebased and LBP-based methods. Comparing the hybrids, our results show that both LBP and oLBP-based hybrids performed better in terms of accuracy than the Eigenfacebased hybrids. In addition, the results also show that the EDA implementation of GEFeS (for the LBP and oLBP hybrids) has the best overall performance. Keywords— Local Binary Pattern (LBP), overlapped Local Binary Pattern (oLBP),Eigenface, Steady State Genetic Algorithm (SSGA), Estimation and Distribution Algorithm (EDA), Feature Selection.