Optimizing Image Feature Selection for Covid-19 Classification Using Bio-Inspired and Meta-Heuristic Algorithms

Abstract

Optimising the selection of optimal image features, particularly for image classification tasks, is a challenging and crucial endeavour. The conventional method of selecting image features independently often results in the selection of unrelated image features, thereby degrading the consistency of classification accuracy. The primary objective of this article is to optimize Meta-heuristic algorithms, specifically Harmony Search (HS) and Tabu Search (TS), by leveraging the capabilities of bio-inspired search algorithms (ACO, BBA, ABC) in conjunction with the wrapper which is a technique that selects a subset of features by evaluating a model's performance using different subsets of features. The essential stages involve adjusting the HS and TS combination with appropriate bio-inspired methods and incorporating the creation of various image feature subsets. Subsequently, a subset evaluation is conducted to confirm the optimum image feature set. The evaluation criteria are based on both the number of image features utilized and the image classification accuracy. Extensive testing has demonstrated that the optimal combination of the selected bio-inspired algorithm and meta-heuristics algorithms, particularly HS and TS, holds the promise of providing an  optimum solution. The solution results in fewer image features and improved classification accuracy for the selected image datasets. Consequently, this research demonstrates that combining bio-inspired algorithms with wrapper methods enhances the efficiency of HS and TS in feature selection.