Opposition-Based Jellyfish Search for Feature Selection

Jellyfish Search (JS) is a recently proposed meta-heuristic optimization algorithm that simulates the behavior of jellyfish searching for food in ocean currents. However, JS suffers from problems related to population diversity in the search space and low convergence rate. This work proposes a new algorithm called opposition-Based Jellyfish Search (OJS), which uses opposition-Based Learning to increase search space coverage and the balance between exploration and exploitation. The OJS is validated against large-scale benchmark optimization functions from the CEC’2013 competition and also against feature selection from six datasets related to fault identification in power transformers. The experimental results demonstrated an increase in the OJS convergence rate concerning the original JS version and a performance improvement, obtaining lower fitness values in the large-scale benchmark optimization functions. Concerning feature selection, OJS obtained better accuracies than JS, demonstrating its viability for identifying faults in power transformers.