Hierarchical-Based Binary Moth Flame Optimization for Feature Extraction in Biomedical Application

Feature extraction is a key challenging task to find the optimal features by alleviating irrelevant features to improve the classification accuracy. The brute force methods yield complete feature space and employ exhaustive search that makes the feature selection a non-deterministic polynomial problem. The meta-heuristic algorithm offers a better optimal solution through random search rather than a complete search. However, the MFO falls with local optima and poor convergence. In this paper, a novel methodology based on the Hierarchical Binary Moth Flame Optimization (HBMFO) with a K-Nearest Neighbor (KNN) classifier is proposed for feature extraction. The motive of this work is to reduce high dimensionality for large datasets through feature extraction. Here, the adaptive hierarchical method has been proposed to update the moth position towards the optimal solution pertaining to searching the space. The simulation results are accomplished on UCI Repository datasets to validate the superiority of the suggested optimization algorithm over existing techniques and feature selection can be carried out with respect to improving stability and accuracy.