<p>Artificial bee colony (ABC) algorithm is a popular swarm intelligence optimization algorithm. When confronted with complex optimization problems, ABC often encounters limitations such as slow convergence speed and low solution accuracy. These issues primarily stem from the algorithm’s emphasis on exploration during the search process at the cost of effective exploitation. Consequently, a dual-population ABC algorithm (called DPEEABC) is proposed from the perspective of balancing exploration and exploitation. For dual-population partitioning, DPEEABC constructs an exploration subpopulation based on diversity and an exploitation subpopulation based on fitness. It also proposes an adaptive population adjustment strategy to ensure a gradual transition in the population structure, shifting from being predominantly exploration-oriented to exploitation-oriented. For dual-population search, DPEEABC integrates the characteristics of exploration and exploitation with the role division among the three types of bees in ABC. Specifically, it designs three customized search equations to highlight the leading role of employed bees in search, the intensified search efforts of onlooker bees, and the supplementary search conducted by scout bees. Experiments on 51 test functions show it outperforms or matches fourteen other algorithms. Real-world applications in mobile robot path planning and esophageal cancer prediction also validate its effectiveness.</p>