NJ-ABCc: A neighborhood- joining artificial bee colony algorithm tested on CEC 2022 and real-world structural design optimizati̇on

This study introduces NJ-ABC, a novel enhancement of the Artificial Bee Colony (ABC) algorithm designed to overcome key limitations of the original approach, including slow convergence, premature stagnation, and weak exploitation. Inspired by the Neighbor-Joining (NJ) technique, NJ-ABC uses a neighborhood-based clustering mechanism to enable ordered solution grouping depending on spatial proximity. While maintaining global variety, this encourages more concentrated investigation inside interesting areas of the search space. The method dynamically changes the exploration–exploitation trade-off using an exponentially declining adaptation factor and uses Latin Hypercube Sampling (LHS) for first population generation to guarantee wide coverage. A PSO-inspired scout bee strategy improves the regeneration of stagnated solutions and gradient-based local search is applied for last refinement of the best solution, so strengthening robustness. Using the CEC 2022 benchmark suite, NJ-ABC showed better convergence accuracy and stability than several state-of- the-art metaheuristics including PSO, TSA, GWO, and advanced ABC variants including MNG-ABC, IGAL-ABC and OCG-ABC, achieving a success rate of 83.33%. Furthermore, the method was effectively used to a practical engineering challenge concerning the cost optimization of reinforced concrete column design under uniaxial bending. Results validate NJ-ABC's performance as a dependable and competitive optimization tool for both theoretical benchmarks and actual structural design problems.