An Improved Binary Grey-Wolf Optimizer With Simulated Annealing for Feature Selection

This paper proposes improvements to the binary grey-wolf optimizer (BGWO) solve feature selection (FS) problem associated with high data dimensionality, irrelevant, noisy, and redundant that will then allow machine learning algorithms attain better classification/clustering accuracy in less training time.We propose three variants of BGWO addition standard variant, applying different transfer functions tackle FS problem. Because generates continuous values needs discrete values, a number V-shaped, S-shaped, U-shaped were investigated for incorporation convert their binary. After investigation, we note performance is affected by function. Then, first look reduce local minima integrating an exploration capability update position grey wolf randomly within search space certain probability; this variant was abbreviated as IBGWO. Consequently, novel mutation strategy proposed select worst wolves population which are updated toward best solution based on probability determine if either or randomly. The selected linearly increased iteration. Finally, combined IBGWO produce second LIBGWO. In last simulated annealing (SA) integrated LIBGWO around best-so-far at end each iteration order identify solutions. validated 32 datasets taken from UCI repository compared six wrapper methods. experiments show superiority improved producing classification than other algorithms.