Hybrid Marine Predator Algorithm with Simulated Annealing for Feature Selection
Utkarsh Mahadeo Khaire, Ramasamy Dhanalakshmi, K. Balakrishnan · 2022
The Marine Predator Algorithm (MPA) is a recently proposed meta-heuristic algorithm based on the survival of the fittest theory. Despite being an outstanding optimization technique, the conventional MPA could not achieve the optimum value of the design parameter due to its premature convergence (PC). The predictive model failed to yield higher accuracy in the process of feature selection from the high-dimensional microarray datasets of life-threatening diseases due to the anomaly of PC in MPA. To address the shortcomings of the existing MPA in feature selection, this paper proposes an improved version of the MPA (iMPA) that uses simulated annealing (SA) to improve its ability to explore a larger design space. The performance of the iMPA is investigated on high-dimensional microarray cancer datasets. The experimental results of the proposed approach outperform conventional MPA optimization techniques and show significant developments in knowledge discovery from high-dimensional microarray datasets.