a modified grey wolf optimizer by individual best memory and penalty factor for sonar and radar dataset classification

Meta-heuristic algorithms (ma) are widely accepted as excellent ways to solve a variety of optimization problems in recent decades. grey wolf optimization (gwo) is a novel metaheuristic algorithm (ma) that has been generated a great deal of research interest due to its advantages such as simple implementation and powerful exploitation. this study proposes a novel gwobased ma and two extra features called individual best memory (ibm) and penalty factor (pf) to train feedforward neural network (fnn) for the classification of sonar and radar datasets. besides, fnn is accompanied by feature selection (fs) using gwo. experiments were done on sonar and radar datasets obtained from the university of california, irvin (uci) to evaluate the performance of the proposed ma; the results demonstrated the proposed ma is markedly better than gwo in terms of classification accuracy, avoiding local optima stagnation, and convergence speed. this framework can be applied to naval navigation systems or atmospheric research.

A Modified Grey Wolf Optimizer by Individual Best Memory and Penalty Factor for Sonar and Radar Dataset Classification

Metaheuristic Algorithms (MA) are widely accepted as excellent ways to solve a variety of optimization problems in recent decades. Grey Wolf Optimization (GWO) is a novel Metaheuristic Algorithm (MA) that has been generated a great deal of research interest due to its advantages such as simple implementation and powerful exploitation. This study proposes a novel GWObased MA and two extra features called Individual Best Memory (IBM) and Penalty Factor (PF) to train Feedforward Neural Network (FNN) for the classification of Sonar and Radar datasets. Besides, FNN is accompanied by Feature Selection (FS) using GWO. Experiments were done on Sonar and Radar datasets obtained from the University of California, Irvin (UCI) to evaluate the performance of the proposed MA; the results demonstrated the proposed MA is markedly better than GWO in terms of classification accuracy, avoiding local optima stagnation, and convergence speed. This framework can be applied to naval navigation systems or atmospheric research.