Akademik Veri Yönetim Sistemi
Neural Computing and Applications, cilt.35, sa.35, ss.24897-24923, 2023 (SCI-Expanded)
Honey formation optimization (HFO) is originally proposed for design problems where the definitions of the objective functions are known priori or under design. HFO extends the Artificial Bee Colony (ABC) algorithm with the concept of multiple components in a source and the worker bees tending to collect components currently needed. However, the necessity of component design for a particular problem makes the HFO not applicable to optimize an arbitrary objective function. In this paper, HFO with single component (HFO-1) is proposed in order to remove this hardship of HFO for numerical function optimizations. Unlike the HFO, which only models the honey formation inside the bee, the HFO-1 further model the honey production process in the hive where sources are turned into honey-forms inside the bee and mature in time through various types of mixing processes using enzymes until the whole mixture becomes mature in the hive. During mixing process, the Pbest (population best) is used as primal catalyzer that metamorphoses other forms towards itself. When the current mixture is mature, a new mixture is started from a new site and saturated with the Gbest (global best) to fasten the maturity of the new mixture towards Gbest . HFO-1 is original in that it extends the formation phase of HFO with novel local search and importantly introduces 3 new phases, mixing, maturation, and saturation, specific to honey production. In this article, 6 algorithms (Whale Optimization, Differential Search, Particle Swarm Optimization, Improved Grey Wolf, Moth-Flame Optimization, HFO-1) are comparatively studied on the basis of 60 popular benchmark functions, containing CEC2019 functions. The results show that HFO-1 is superior to others according to mean absolute error, mean variance and Wilcoxon Rank-Sum Test analysis.