Randomized Shuffled Hierarchical Partitioning Technique for Enhancing Efficiency of Swarm Algorithms

Hierarchical partitioning (HIER) is an efficient multi-population partitioning technique. By enhancing swarm algorithms’ exploring capabilities, it improves their efficiency and aids them in evading local optima. Modified hierarchical partitioning (mHIER) was recently proposed to reduce the duplicity in HIER and guide the solutions toward better regions. In this paper, three modifications have been proposed to enhance the efficiency of mHIER. The first modification is shuffling of sub-populations, which facilitates better information exchange across the entire population. The second modification is randomization which is introduced to help counter duplicity. It works by adding a new randomly generated solution every time local solutions are exchanged. The third approach, randomized shuffled hierarchical technique (RSHier) is a combination of these two. The three techniques are compared over bat algorithm, and RSHier is found to be the best. To establish the effectiveness of RSHier, it is further tested over four additional swarm algorithms: firefly algorithm, flower pollination algorithm, moth search algorithm, and particle swarm optimization algorithm. Experimentation is done over 30 benchmark functions and CEC 2014 function set. Computational results establish RSHier as an efficient population partitioning technique to enhance the efficiency of different swarm algorithms.