Binary Grey Wolf Optimizer with K-Nearest Neighbor classifier for Feature Selection

Abstract

Iteration number and population size are two key factors that influence the effectiveness of a certain feature selection algorithm. Randomly choosing these factors, however, might be an impractical approach that could lead to low algorithm accuracy. In this paper, we assessed the changes in the accuracy of Binary Grey Wolf Optimizer (BGWO) at varying a function of iteration number (50,100,150 and 200) and population size (10,20,30) in four benchmark datasets. The results generally indicate that there is an optimum iteration number (T) beyond which the accuracy of BGWO started to decrease. Similarly, it was seen that an optimum population size (N) exists, which yield a high average accuracy of the BGWO algorithm. The findings suggest that it is essential to optimize the iteration number and population size before the execution of BGWO.