Artificial Bee Hive Algorithm: Agent Behaviors and Benchmark Trade-Offs

Summary

This article continues the implementation of the Artificial Bee Hive Algorithm, a population-based optimization method inspired by bee foraging. Agents occupy different behavioral states, including novice, experienced, search, and food-source roles. Their movements combine random exploration, following information from other agents, directed movement, and local search around an agent’s best known position. Experienced agents select among actions probabilistically, while novices choose between random search and following other agents.

The article evaluates the completed algorithm on test functions and summarizes its observed strengths and weaknesses. It reports good results on low-dimensional and discrete problems, but weaker convergence on high-dimensional smooth functions. The method also has complex logic and many external parameters, which may complicate implementation and tuning. These are benchmark findings for an optimization algorithm, not evidence of trading performance; applying it to strategy search would require separate validation against relevant data and baselines.

Key ideas

• The algorithm represents each bee as an agent whose behavior depends on its state and available information.
• Novice agents choose between random exploration and following experienced agents.
• Experienced agents can also search near their best known position, while search and source states use distinct movement rules.
• The reported tests found strengths on low-dimensional and discrete functions, alongside weaker convergence on high-dimensional smooth functions.
• The algorithm’s complex logic and large number of external parameters are practical drawbacks.

Tags

This summary was written by Stratmill's research agent from the original; it is not a copy of the source.