A Systematic Survey on Foundational Swarm Intelligence: Comparative Mechanics, Applications, and Limitations of ACO and BCO

Abstract

Nowadays, swarm intelligence (SI) has become a significant concept in artificial intelligence. It refers to the idea that the collective behavior of self-directed systems can be used to solve challenging NP-hard optimization problems. A variety of research has been conducted recently to develop more efficient SI algorithms, and new metaheuristics have been introduced to handle distinct optimization tasks. However, most of the research effort is concentrated on ant colony optimization (ACO) and bee colony optimization (BCO). This article aims to provide an overview of these two approaches to make a comprehensive comparison between them. We focus on ACO’s pheromone-based stigmergy and BCO’s division of labor and recruitment mechanisms. Furthermore, we provide a detailed theoretical analysis of both algorithms. We compare their convergence, performance, and complexity and evaluate their applications and performance in practice (e.g., ACO-tagger demonstrated 96.87 % accuracy in Part-of-Speech tagging and ABC-RNN achieved 98.6% performance in financial fraud detection). Finally, we discuss their limitations and directions for future research to advance swarm intelligence research, including hybrid approaches that combine classic SI algorithms with state-of-the-art methods in machine learning.