A new hybrid model of chaotic Metaheuristic optimizer for feature selection

Abstract

Recently, there have been many efforts to invent a new ideal model for feature selection in big data applications using multiple approaches including

the most important approaches which are those based on Metaheuristics algorithms, in order to solve the problem of the high complexity that faces algorithms applied to big and massive data sets.

Metaheuristics algorithms have made great progress in many fields related to optimization (obtaining the optimal solutions (maximum - minimum)), as they can generate good solutions in reasonable periods of time, but on the other hand they suffer from many disadvantages such as falling into the local optimal solution trap (stuck at local edges), lack of search diversity (occupying the entire search space by search agents) and imbalance between Exploitative and exploratory performance.

In this paper a new hybrid binary feature selection model called “Chaotic Salp Swarm Algorithm Whale Optimization Algorithm” (CSSAWOA) was proposed to solve Multi-objective optimization problems, where the basic idea of CSSAWOA is to improve the Salp Swarm Optimization Algorithm (SSA) by hybridizing it with the Humpback Whale Optimization Algorithm (WOA). In order to share their strengths.

results obtained from applying the proposed model were compared with results obtained from applying some famous and original Metaheuristics algorithms as: Particle Swarm Optimization (PSO), Gray Wolf (GWO), Salp Swarm (SSA), and Humpback Whales (WOA), using /15/ diverse data sets in size, big, medium and small, according to the following criteria: Obtaining the least number of relevant features - Increasing classification accuracy - Reducing processing and computation time.